>> From the Library of
Congress in Washington, DC.
>> DAME WENDY HALL: Thank you
for finding your way back again,
I hope you had a good something
to eat and a breath of fresh air.
There will be two panels
this afternoon.
The first one is putting
data for work.
For me, and I'm going to
talk more about this later,
the key to helping
people find the archives
that are around is all the data.
And that's what this panel is about.
Focusing on types of data, how you
produce it, what you do with it
and some of the ethical questions
of handling this type of data,
which is very different
to handling data
from the physical sciences
for example.
Now, you have the bios of
everybody on the panel.
So, I'm not going to
introduce people at length.
I'm going to take people in the
order that they're on this sheet.
And luckily for me everyone
on the panel is on this sheet.
So, we're okay [laugther].
So, we're going to start with Lee
Rainie who's infamous, no famous,
sorry for being the
Director of Internet Science
and Technology Research at
the Pew Research Center.
And we first met in Hong Kong.
I've got that right
this time, haven't I?
Yeah, and this was,
oh a year or more ago.
And I said, oh I'm
coming to Washington.
Oh, let's do something together.
And here we are.
So, Lee you have your
time in the limelight now.
>> LEE RAINIE: Thank you very much
and thank you for doing this Wendy,
it's an amazing resource and if the
Doris Kerns-Goodwin of 100 years
from now is a good enough historian
she will discover what an important
event this was and why
it led to the good things
that she was writing
out in the next century.
And anytime you can be in
the same room with Vint Cerf,
first of all, you have
to say thanks.
And so, we all say thanks to him.
I just wanted to start by underlying
a point that Richard Marciano made
about how data quality now
is so different from before
and it's the phase change in data
that really makes this a different
kind of environment for us.
So, everybody talks about
volume, but one of the artifacts
of greater volume is that just
more people get to participate.
One of the striking things we see
in the big national surveys we do
about the impact of the internet
on people is how grateful they are
that they can tell
their stories now.
And they're not mediated through
industrial era media companies.
And they get to tell their
own stories in their own way.
And they get to find other people
who share those stories
and those circumstances.
And it's an enormous impact.
They're annoyed at the millions
and billions of other people
who tell their stories,
especially the wrong ones,
or the ones that are
spun up or lied up.
But, by and large when people are
assessing their own information
environment, they're so grateful
now that they have a voice
and they have place in way that
you know GeoCities gave them a long
while back.
The second dimension,
so volume is one.
Velocity is the other.
So, it's not just the pace
of information that's coming
through sort of normal channels,
but it's real time information.
You know we can now
talk about location
as a really important indicator and
predictor of people's activities
and behaviors and things like that
in a way we never could before.
You know Pew is a big
survey organization
and people are not very reliable
when you ask them specific
questions about their activities.
Their media behaviors, their
location behaviors, their frequency
of doing things and stuff.
Well, now we have reality
checks on that
when we have real time information
coming at the speed that it does
and being manipulated in
ways that can send signals
about what they're doing.
And the variety of
information is also changing.
Sort of in the same way that
people are glad to be able
to tell their stories there
are so many other places now
that can tell their
truth to the world.
And this is enriching the amount
of information that's available.
It's certainly expanding the number
of gatekeepers of information.
It's changing the relationship
that people have with expertise.
Who's an expert, and
who do you rely on
and where do you allocate your
trust and things like that.
Which brings me to one of the
three points I want to make.
I'm going to make three points and
two please in my little time here.
Insofar as you have to make choices
about what information to archive
and what to focus on
and stuff like that,
I would argue that let's
be selfish about it.
Let's tell what we
think, those of us
in the room think might be the
biggest story certainly of our era,
and maybe for a long period of time.
It's the rise of the
internet and the web itself.
It is changing the way people
think about their social worlds.
It's changing the nature
of their social worlds.
It's changing how they're
in community.
It's changing how they
think about expertise.
It's changing their
relationship to big institutions,
little institutions and
everything in between.
It's changing the way they
allocate their attention.
It's changing the way
they can act as agents
in their own behalf economically,
socially and politically.
So, across the board.
We're in the midst of a huge story.
And I'm trying to think of another
time in history where people were
so aware that the ground was
changing underneath them.
Maybe the arrival of atomic weapons
might be something that generated
as much, sort of immediate
commentary
about its impact on humanity.
But that's the only
other one I can think of.
So, we've got this big story to
tell and why not be intentional
in the data that we collect and the
data that we preserve for the future
in telling the story of this era and
all fo the change that it initiated.
You know, in a way that
justifies archiving lots of stuff,
but maybe as you're sort of sitting
down, figuring out what collections
to amass, that could be one filter
through which you pass
your judgement.
The second thing is that ironically
even as the volume velocity
and variety of information is
exploding there is an information
market that is degrading.
And archivists can really play
a big role in helping undergird
that market and it's the
civic information market.
One of the sort of great things
of the industrial era was
that newspapers arose and
newspapers were awesomely good
at bundling together all
kind of different content.
But serving it up to you in a
way where the mayor got covered,
the school board got covered,
the zoning board got covered,
the state house got covered.
National and international
capitals got covered.
Well, as that bundle has been blown
up by the internet and each sort
of constituent part of the bundle
has found its own business model
and its own advertising base, what
has been orphaned in that process
in some sense is civic information.
So, you don't get as much
information about the mayor's office
and you don't find out what's
going on on the school board.
And you don't necessarily know what
the zoning board is up to in a way
that you did when there
was a reporter in the room
when all those things were happening
and when the material
is being released.
And so, archivists can maybe
pay a role in sort of uplifting
that segment, even as the business
model for it is in trouble.
I was so struck, maybe these
are the marching orders
to the datathon people, but I was
so struck that all of them marched
to civic information and
civic insight as a thing
that they thought was so
precious and so interesting
and so easily demonstrable there.
Yeah for you guys.
And everybody that's
part of that team.
And that instinct to fill
in that information gap is
really good and important one.
And finally, there's a movement
that's just poised to help you
in the archiving process and
in the explication process,
which is the open data movement.
You know it's hard to archive stuff.
It's hard to sort of think through
some of the ethical decisions
about scraping stuff
and things like that,
but this is all ours, you know?
The open data movement is built
around information we paid for
and whose copyright status and
whose ethical status is not nearly
in question to the degree that
the proprietary information
or other information is.
And these folks are you know
in the same tribe as you guys.
They love to hack.
They want to figure out how
to make this information more
useful for civic purposes.
So, those are my three points.
Organize around the big
story, the rise of the web
and what it's meant to us all.
Help a degrading information
environment
and with the collapse of newspapers.
And give a hand to the
open data movement.
Now I have two pleas to end with.
One plea is an affirmative one.
Another ally group for you
guys is public librarians.
We do a lot of work related to
public libraries and they are poised
to help, especially in the model
that Vint was articulating
about, distributive archiving.
They kind of have the skillset
and interest all ready.
But they need marching orders
and they need help from you.
And I think that you could
take advantage of that.
They also, when I go out and
give speeches to librarians,
we do a lot of research about
role of public libraries
in people's lives and
in community's lives.
One of the things I say that
librarians can do in the future.
There would be a real
service that sort
of matches an industrial age
service that they provided,
is the algorithm fact checkers.
They need training in that, they
need to know how to do that.
But one of the things that as we
move into the age of the internet
of things and big data
analytics and stuff like that.
Somebody's got to watch
the watchers.
And the wonderful computer
scientists in the room are going
to do a good job at that.
But there's an army out
there that might be willing
to help you on that [laughter].
And the final.
So, think of public librarians
as good, useful, ready to charge
into battle allies for you.
And the final plea I'll make is
that even as all of these data arise
and the exciting ways to think
about capturing it and analyzing it
and stuff like that, don't
make the mistake of thinking
that it represents everything
and everyone who's important.
Twenty percent of American adults,
one-fifth of America
adults use Twitter.
Great resource, wonderful
to analyze,
fabulous to do social
networking analysis off.
But you know four-fifth of
adults don't use it in America
and the number is even higher
in other developed countries.
And even higher in
developing countries.
It's you know and even for Facebook,
1/3 of American adults
don't use Facebook.
So, don't necessarily think that
you are telling the full, full story
and it's certainly a representative
story by thinking that all
of this stuff that you're
amassing and analyzing stands
in for everyone, because it doesn't.
and there are ways that you can
backfill for some of those issues.
And there, you certainly,
sort of every day you get
up you should be thinking
about how can I serve
to narrow digital divides
rather than expand them.
Thanks.
[ Applause ]
>> DAME WENDY HALL:
That was wonderful.
And you're very neat, you kept time.
And I'm fascinated by
how we move particularly
with the elections coming up, well
our referendum and your election,
which all seem completely
crazy to me.
I wish we weren't having
a referendum.
It's driving us apart and
it didn't need to happen.
But the how we go from you know
the century we've had of polling
to the century of social media and
how these, you know how we do begin
to forecast what the
people are thinking.
Now, it's my great pleasure
to introduce someone
who has become a great friend
of mine Professor Katy Börner,
who is a professor
Indiana University.
You have the list of the
things that she does there.
She's one of the most
amazing people.
She's a force of nature.
She produces the most amazing
maps and I know she's going
to show you this, but she produces
things like this, sorry, this here,
this amazing book of
maps, of science.
And she's also one of the most
generous persons I've met in terms
of her time and her wish
to work with others.
Kathy, over to you.
>> KATY BÖRNER: Thank you, Wendy.
And thanks to all the
organizers for hosting us
in this beautiful palace
of knowledge and wisdom.
Forward. So, I wanted to start
with a suggestion to chart the web,
to map it, to create an atlas
of all the different ways
that this web has been
helping us to find our way
to understand structures and
trends, to identify problems
with the data that exists, truly.
To also understand what data,
we don't have because just
geospatial regions which are black
if you overlaid a number
of holes there.
And to then take these maps
and bundle them all together
in what is known as an atlas.
And the presentation will start
with a few maps that are part
of the mapping science
exhibit and that show
that not only science today,
but also [inaudible] truly
connects us globally.
And the first map I'm
showing here is a map
of scientific collaborations
between 2005 and 2009.
Some of you might have
seen a similar map.
On Facebook linkages.
And so, this is I think
inspired by that map.
And it really shows that we, today
are very much globally connected,
but you also get to see
that there are some areas
which are densely interlinked
and others
which are more sparsely interlinked.
I also saw this recent map which
some of you might have seen
in the news media and it
shows a cartogram style map
of how many domain
names actually exist
and how many different people
use these different domain names.
Another map, which was actually hand
drawn is shown here and you can zoom
in to see the big Google
landscape and data oceans etcetera.
All these maps, by the
way are available online.
So, you can zoom into them.
So, if you go to simaps.org
you will find a hundred maps
that have been created by
more than 250 scholars.
This is a map of Twitter activity
and you see that in the Netherlands
and you might also know
that in the Netherlands,
people are holding a drink in
one hand, and their texting
with the other hand,
while they're driving.
[ Laughter ]
But you also get to
see that these maps
of course redraw major
capitals, major railway
and other transportation lines.
And you also get to see what
areas in Europe are multi-lingual
and what kind of languages
that might be.
So, I think the plea here to
take the data we have and to try
to not only explore that data,
but also to communicate it
to a much larger audience.
In an attempt to communicate
what we know about the work,
but also to communicate what we
don't know where the monsters are.
There's a darkness's are.
A team which also includes
some of my colleagues
at Indiana University
has used Twitter data
to map the mood of a nation.
And it's interesting to see how
grumpy some are in the morning
and how happy they are
when the sun gets up.
And I won't have time to look at
all these maps in detail here.
But if you're interested in maps,
there's some handouts in the front
and again they're all online.
This is a map which is
near and dear to my heart.
It shows not only how
science sites itself,
which is how typically these
maps of science are created,
but how it's used, for instance in
doctor's offices by practitioners
which are desperate to have
parents go home with their child
and not without their children.
And so, here you get to see how
these different scientific papers
are downloaded one, and then
the next, and then the next.
And then there is a sequence between
them, they are connected by a links.
And the more often they're
downloaded after each other,
the stronger the linkages are.
And the more closer in space they
are given these layered algorithms
we are using here.
Now, it would be wonderful if all of
us could actually read those maps,
and could make sense of them,
and could find our ways not only
in geospatial space, but
also in knowledge space.
Unfortunately, however,
this might not be the case.
So, we recently did a study on
six science museums in the US.
It's still my hope that it's
different in Europe or Asia,
but in the US, we asked a thousand
youths and their caregivers
to read 20 maps, or each one
of them got 5 maps to read.
And it turns out that those which
make it into science museums,
they actually do have a hard
time reading scatter plots,
reading cartogram maps,
reading networks.
Networks are almost impossible
for many, many out there.
So, there is a huge need to help
people not only read and write text,
but also help them
read and write data.
And I would like to make the
argument that if you only read text,
but never write text, then it's very
hard to actually be a good reader.
And similarly, here, if you
only read data visualizations
which are shown in newspapers
or textbooks, but you never get
to make one, then it's also very,
very hard to enter
practice specialization.
So, we have been trying to bring
materials online that empower anyone
from high school to
beyond 100 years of age.
We have some of those in
our MOOC, to understand how
to actually take a dataset and
analyze it and visualize it
and communicate that
result to others.
And we use visualization framework,
which first of all
introduces different task types
from statistical to
temporal, topical,
geospatial and network analysis.
And different levels of analysis.
So, from micro individual
level studies, to meso,
to macro population level studies.
And if you now have that
table, which you see here,
then each cell actually refers
to a set of tools that are meant
for that particular purpose that
are developed by geographers,
or cartographers if it's
a rare question you have,
or by network scientists if it's
a network, with whom question.
Or by linguist if it's
a what question.
So, these tools are really developed
in very different areas of science,
but oftentimes you might have
multiple of these questions.
So, this helps you to kind
of maneuver the landscapes
of different tools and workflows.
And then, this is a
very simple workflow
of how you can work
data into knowledge.
And it's always one, two, three.
Because it has to be so simple.
You read data, you analyze
it, you visualize it.
And then you select a reference
system, could be just two axes
or a map of the US for instance.
Then you overlay the data and
then you start encoding the data.
And then you will probably see
that you don't have the right data,
or the data has a problem,
and then you do it all again.
So, it's a very iterative process.
And again, these are just different
reference systems you could swap
out, and this is a course schedule.
And there are two books,
one if more timely,
and it's actually co-written
with a librarian.
So, Ted Polly he now is working
in Indianapolis as a librarian.
And he has been instrumental
in bringing these materials
to many libraries in
the US, but also abroad.
And then "The Atlas of
Knowledge" is a collection
of these different
maps has good examples,
but also has the visualization
framework.
And we also use our
own microscope tools.
And apparently, I don't have
any time to go into that detail.
But ultimately students
are supposed to understand
that if they have an Excel file
with different columns then they can
point to different types of columns.
Let's say the year and topics.
And then they can run a
topical analysis of bursts
of activity using these
two columns as input.
Or if they have a column with
addresses and they can geolocate
and they can overlay it over a map
of the world, or a map of the US.
And just by helping them understand
what columns a normal Excel file
facilitates what kind of analysis
they are far a step ahead.
And then the tools also use that
organization in the menu system.
And I'm out of time.
Okay, so the visualization
framework can be found in that book
if you like, but what I wanted
to do it, I'm out of time,
make sure that you
also take home a flyer.
So, maybe while I distract Wendy.
Yeah, yeah.
I'm not going to do it.
I am out of time.
I don't know why I
wanted so many slides.
I'm very, very sorry [laugther].
>> DAME WENDY HALL: This is
a typical academic all right.
>> KATY BÖRNER: Found
my slide, sorry.
So, those of you which do not
only want to look backwards,
but want to look forward
they might like to check
out that recent conference,
which Wendy was also involved
in as a keynote speaker.
Because now all the slides and
all the recordings are online.
So, if you want to also use data to,
for a certain degree
predict the future,
I would recommend that website.
Sorry.
[ Applause ]
>> DAME WENDY HALL: I love the
way we academics supposedly have a
different view of the
space time continuum
to everybody else [laugther].
But I would recommend you
to look at this material.
Delve into those websites.
And some of the students from here
are going to visit you next week.
So, they'll be playing
with all this data.
And the conference as well was
a really brilliant get together,
everybody.
Now, it's my please to
introduce someone I've work
with for a long time.
A very good friend of mine,
Professor James Hendler.
Lee talked about open data, and
of course Jim, who is at RPI,
he's very well known for the work
he's done on data at data.gov.
And also, his seminal work
with Tim Berners-Lee and others
on the semantic web, which I think
is really important in this world.
This is not a conference
about the semantic web,
but maybe Tim will,
Jim will touch on it.
Sorry, slip of the tongue there.
And he's also the chair
of the website's trust
so we work together on that as well.
Jim, over to you.
>> JAMES HENDLER: Thank you Wendy,
and behalf of the Web Science
Trust let me welcome everyone.
We're very excited to be helping to
sponsor this event to some degree.
And I'll let it go at
that because I don't want
to use all my time on that.
But, I also was going
to take some time
to refute the most controversial
claim made this morning,
which was that the
internet was about puppies.
But I don't have time for that.
Just Google Tim Berners-Lee and cats
and you'll find the correct answer.
But what I'd really like to do today
is start with a thought experiment.
We're all sitting here in
the Library of Congress
and particularly those of you who
are librarians and library trained,
you won't need much thought about
this, but for everyone else,
imagine if when this building
was first being conceived,
when this concept was first being
conceived, we had a search engine.
That was a keyword
based search engine
like what we have on the web today.
Right? So, instead of putting these
books in here, they were just thrown
in kind of at random, each
book was just given a number
and you would use the
search engine, right?
That would have been a
terrible way to build a library.
Right? and 140 years ago, this year,
the person who every librarian sort
of loves to hate and hates
to love is Melvil Dewey came
up with a very famous system;
the Dewey Decimal System.
Which sort of way a way of
taxonomizing the whole world.
But it put books near each other
if they were sort of
conceptually close.
And again, we can argue lots
of things there've been many,
many attempts over the years
to improve on that system,
and yet it still hangs on.
Partly for historic reasons,
but partly because it's
hard to do it differently.
The problem is for data, that's not
even the word search engine isn't a
viable approach.
Right? The umber 13
sitting in a column
in my database could be anything.
Could be somebody's age,
could be representative
of some set of columns, anything.
So, one of the real problems we
have with data, in fact with all
of the indexing of the web, but
particularly the data on the web,
is that we need an index structure.
We need that Dewey Decimal
System replacement; however,
it clearly won't work to do
something very similar to Dewey.
So, what I want to sort of hit on
today is four things we really need
to think about in terms of
organizing the web archive,
but particularly the
data aspects of it.
So, for the web itself the URL
system provides at least some kind
of organizational structure.
And for the internet archive that's
to a large degree what's been used.
And then on top of it you just try
to build very topical
things, and that works.
But again, in the data world,
in the next talk we'll talk more
about linking data
so I don't have to.
But linked data is happening,
but that's not sufficient,
because again, that's about how to
get data organized in certain ways,
but it still doesn't tell us
how to find it, so discovery.
It doesn't tell us how to integrate
it doesn't tell us whether what this
person was using here is similar
to what this person is using there.
So, some observations, you know
sort of based on those 140 years
versus what we need to do.
So, the first thing, clearly
anything we're going to do
about web data can't
be in a taxonomy.
It can't be a tree.
Vint put it very beautifully, he
said, the web is a web, right?
And it's not even just a graph.
It's a complex hypergraph structure
with a lot of different connectivity
and a lot of different kinds.
And the data that lives and
powers that also tends to live
in that representational scheme.
So, number one is how
do we even begin
to build something that's
an organizational structure
that doesn't live in the kind
of trees that we're taught
in library school is the
best way to organize things.
The second thing is
we really cannot name
down to the individual
data elements, right?
Really a lot of the data
sharing is about metadata.
And you've heard that
word a lot today.
And I'm not going to spend the
four or five hours I'd love
to spend explaining it in detail.
But, we really need to
be thinking a lot about,
there's this data here, what is it?
What is it about?
How do we find it?
So, once we get there, a lot
of techniques can be used
to figure out what's in it.
But the real question is how do
I know and a lot of the things
that I've been trying to do
learning about data bottom
up are missing this point.
So, having figure out
that this is a name
and that this column
is a data, right?
Makes a huge difference
if I'm reading a mortality
database, or a maternity database.
Right? One is the day
somebody was born,
the other is the day someone died.
Right? But there's nothing in
the database that tells you this.
It's in the description of
what the database is about.
So, the second thing is we really
need to figure out ways to talk
about the datasets and how
they change and what's in them.
The third one I'm going
to call more locality.
And this really talks to
that question we were asked
about cultures, languages, and
I'm going to add context to that.
So, we need linking across those
things, but we also need the ability
to find things across these
different ways of thinking about it.
We're going to see
something a little later on,
I think Wendy's going to talk
about a project called
the Web Observatory,
which we've been involved in
together, but one of the problems
in that is again we're still
organizing the data based on sort
of the simple terms that
are put in by the creators
and we don't have a way of saying
these people who put it he dataset
in Chan Hi are thinking about
the communication differently
than these people who put
it in in San Francisco.
Even though there may be
shared terms or context.
And finally, we need a temporal
model, because data is not static.
Books change slowly.
And you know it's fairly easy
to refer to the second printing
of something, or the
second edition of something.
Very hard to talk about this
particular data that was pulled
on such a such a date
at such and such a time,
where it is changing
at this kind of speed.
So, when we were doing the
open data for data.gov one
of the big issues we were playing
with was how do we tell people
how often a given dataset changes?
And the decision was made
at a higher level than mine
to ignore that issue, right?
Because from the user's point of
view it didn't really matter much.
It was just the developer
making sure they were pointing
at the most up-to-date API.
But from the archivist view,
that's crucial if we want
to know what the data was
when we have to have a way
of talking about that when.
There are some other but I
think those are the big four;
that we need a way of
doing it not as a taxonomy.
We need something that
goes to the metadata level,
rather than to the
data element level.
We need some way of linking across
localities, cultures, and things.
And we need a way of
capturing the temporal
and changing nature of data.
So, actually when the sematic web
was first created there was a view
of building information
sources that live near the data,
linking them together and
having that grow bottom up.
And what sort of has happened
over time is what's been called
link data has sort of moved
down to just trying to put data
elements together and that sort
of organizational structure has
moved up to languages like Owl
and complex ontology languages,
which have a role, but
it's not this role.
So, a lot of what I'm pushing
nowadays and starting to try
to get a conversation and
really find the place to hold
that conversation, is how do we
start to organize the information
about discovering data,
integrating data,
once we've pulled it together
validating data, some of the stuff
that would then feed
explaining and exploring data,
the [inaudible] etcetera.
So, I'll stop there, but
we really need new models
of web ontology, of
link data etcetera.
That are tied to data description
linking ingest, integration
and all those things Vint was
talking about today so that
when we go looking for
it, we can find it.
I'll stop there, thank you.
[ Applause ]
>> DAME WENDY HALL: Jim, you're
my most disciplined panelist
so far, well done.
And that was brilliant.
Thank you.
And our last panelist
for this panel is Philip,
I can't pronounce your name
you've got too many consonants.
>> PHILIP E. SCHREUR: Schreur.
>> DAME WENDY HALL: Schreur.
I've tried, I've practiced and
practiced and completely failed.
So, there you go.
Philip's come all the way from
Stanford to be with us today.
He's the assistance
university librarian there.
He's responsible for
technical and access service
in the library at Stanford.
Which is so crucial to what
we're talking about here.
So, we're really grateful that
he's come all this way to talk.
And I'm very interested to see he
has a PhD in medieval music theory.
We're getting quite a lot of
this analogies with the past
to where we are today in terms
of we think we're struggling
and it is a very different
environment.
So, we see the impact of what
we're doing as we do it almost.
But people have struggled
with all sorts of ideas to do
with knowledge in the past.
And so, there's nothing new there.
It's maybe just scale.
But the other thing I
do know about Stanford,
because Mike Keller who's the heal
librarian there is a good friend
of mine.
And they have done amazing work
on the semantic web
and digital libraries.
I don't know if Philip is
going to talk about that today,
but they are one of
the leading libraries
in the world in that respect.
So, Philip, over to you.
>> PHILIP E. SCHREUR:
So, thank you very much.
I'm here to talk about the topic
today really from the point of view
of an academic research library.
So, the first thing I wanted to say
was that I am very fortunate to work
for a university librarian
that likes to throw
out almost impossible challenges
with a certain amount of regularity.
So, I remember back in the late
nineties when I was the head
of the metadata department
at Stanford,
he said could you provide
intellectual access to all
of the internet and bring
it into our catalog?
So, I have been puzzling over that
for the past, almost 20 years.
And not in particular relation
to that, but I resigned
that position shortly thereafter.
Although I did come
back to Stanford.
More recently we have been
very focused on trying
to shift the basic technology
about the way we do our
business towards link data.
And then in a conversation with Mike
six or seven years ago, he said,
once we make this transition I don't
want you to provide access simply
to the resources that the
library traditionally holds,
I want you to produce
access to all of the data
that the university generates.
So, another nearly impossible task.
But I think both thoughts
are very pertinent
to what we are talking about today.
I think for us in the library key
to putting the data work
is discovering the data
in the first place.
So, in my case that would mean
either bringing all of the data
and metadata that's on
the web into our catalog,
which I think is both
impractical and impossible.
Or, it means putting our data out
onto the web, which is something
that we are working on right now.
Or, it means providing some sort
of federated search between what is
in our catalog and
what is out on the web.
So, unfortunately, the library
environment is a very closed one
right now.
It is really focused on metadata;
it's not focused on data.
But it's also focused on metadata
primarily in the Marc formats.
Which were wonderful in their time,
but they were developed in the 1960s
in order to represent
metadata that was printed
or typed on library cards.
So, when you think about trying
to represent the data on the web,
in the Marc formats, thinking
about them as catalog cards,
it's not a very successful
transition.
So, we really needed to
move onto something else.
And I think link data
will be that one.
So, as I said for the past number
of years there have been a lot
of efforts in libraries to transform
our data into link data to be able
to get it out onto the web where we
see people are searching for data,
to be able to incorporate what we do
with what is out there on the web.
Right now, we are focused
on an effort
to change our basic functionality,
the way we create data,
to create data natively
in RDF instead of focusing
on creation of Marc and conversion.
Because we've discovered that
that conversion process is
extremely lossy.
Marc was developed trying
to reproduce a catalog card,
so when you saw the
data on the screen,
it's as if it were a catalog card.
And there were a lot of
relationships which is left
up to you in your own
mind to create.
It's not coded there in the
data, so when you convert it
to link data none of those
links are actually there.
But as I said, it's turning out to
be a very costly transition for us.
It's not only a particular
library that needs
to change the way it works.
We are in our own web
of data suppliers.
We have data suppliers who give
us cataloging that give it to us
in Marc, we communicate with
other libraries nationally
and internationally in Marc.
Our ILS system by which we
do things like circulation
and payments, all the
data is in Marc.
So, all of that needs to change
as an entire ecosystem before we can
successfully make that transition
and it's going to be a
lengthy and expensive one.
As I thought about hoe libraries
need to make use of the data that's
out on the web, I think one of
the biggest challenges for us is
that the data are so
extremely variable out there.
One of the things which gave us
most consternation recently was
discovering that the
Bulgarian census is now put
out as a web document, as opposed
to a traditionally printed document.
So, that caused a lot of problems
for us as far as archiving went.
But again, it's a great example
of a resource on the web,
which really closely mimics the
traditional library resource.
And in those cases, we
would like to bring it in,
give it our traditional cataloging,
to mix it with the resources
that it's very closely related to.
So, that is one thing.
But on the other hand,
there's a very small percentage
of internet resources, which
are really of that type.
And so, we are simply
overwhelmed by those numbers.
Our current model of using humans
to provide intellectual access
to individual resources, simply
will not scale to the web,
it's an impossible situation.
So, even if we shift to link data by
which we could incorporate that data
into the web, we could
never describe the web.
So, we're still left
with a conundrum.
I think the other thing is
that this symposium really is
about making the data
work, it really wasn't
about making the metadata work.
Although metadata has been
mentioned a number of times.
So, when I think metadata,
it is a type of data,
but it's really data
with an ulterior motive.
You know it is sneakily trying
to link things together in a way
that you can make them discoverable.
But often it does it in a
way, it loses some specificity
of description in order to do
that linking across domains.
So, it is a very tricky
concept to be able to work with.
So, I think if we really
are going to try
to incorporate the data that's out
on the web with the data that's
in traditional library resources,
we are going to have to focus
on a much more automated process
for providing access
to that information.
So, in dealing with sort of
more traditional web resources,
look at things like automated
entity extraction in assignment
of identifiers and
reconciliation behind the scenes
to be able to draw things together.
Also, I think we need to look
at some things automated
semantic analysis of content
and use a common vocabulary so
that we can draw things together.
For about five years I worked at a
publisher called High Wire Press.
And there we were very interested in
doing the semantic analysis of text.
And it ended up at the end
of the five-year period some
of our major subscribers to the
service liked the semantic analysis
of text better than the human
semantic analysis of the text,
because they said, well you
know you both make mistakes,
but at least the machine
makes a consistent mistake.
So, they found it easier
to deal with.
And at least it's high quality.
So, just in summary
when I was thinking
about what's really important for
academic libraries from our point
of view what will be our
greatest challenges in trying
to make the data from the web work?
So, the first one is how we select
which portions of the
web to capture.
So, I know we've talked
about it a lot here.
I think it will come down to
either we divvy up the web,
and decide on certain
parts to collect,
or our collection development
people look at what makes most sense
for Stanford and decide to focus
on those types of resources.
But then, once that's done,
we really need to be able
to find an automated way of
extracting those resources,
which really kind of mimic our
more traditional library resources.
Because we really want to provide
access to them in the way we have
to be able to integrate those
more easily in our online catalog
in that traditional sense.
As I said, for the rest of the
materials, we need to focus
on that sort of more automated
ways of providing access
and providing metadata
through automated entity
extraction and semantic analysis.
A huge problem, which we're
only beginning to deal
with is the assignment
of identifiers
and their reconciliation.
For integration for
discovery purposes
across the web in the
library catalog.
We know link data sounds intriguing
and it sounds like a great solution,
but you know worldwide there
will be many identifiers assigned
for that same entity.
And we need to find a way, and an
automated way of reconciling all
of that otherwise nothing is linked.
And I think the last thing
that we've been puzzling over,
and I really don't
know the answer to.
We really want to be
able to make sure
that those semantic associations
that we create persist over time.
So, we've talked a lot
about preservation here,
but I think it's a
different concept.
It's persistence.
It's not preservation.
And I think it's one of those topics
that we're just really
beginning to struggle with now.
Yep.
[ Applause ]
>> DAME WENDY HALL: All right.
We've got plenty of time.
Well done panel.
And I could sit and listen
to you forever Philip
because you talk my language
and now I know why I love
working with Stanford Library.
Because they get the semantics bit.
Yeah? And that's a message
to all librarians here.
At some point, you've got
to deal with that properly.
Okay. So, it's hard to see from
here isn't it, where the mics are.
Where are the mics?
Ah-ha. There's a man
with two mics over there.
Okay, who wants to take the
conversation forward then?
You look so tired, some on
this is great stuff here.
Questions, answers.
>> KATY BÖRNER: I've
got a question.
>> DAME WENDY HALL: Oh, let,
Professor Jain, here
in the front row.
You save it, save it.
What I couldn't see, didn't see it.
Here's, right?
>> RAMESH JAIN: Hello
this is Ramesh Jain.
I've been hearing a lot about
data and metadata and when we talk
about metadata, maybe
this is my perception,
but I am getting impression
that we consider
that it's second class
data, is that right?
>> DAME WENDY HALL: Do you
want to take that first, Phil,
because you brought up metadata.
>> PHILIP E. SCHREUR: Right.
I did. So, I wouldn't
say it's second class.
But it certainly is different.
I mean we had a conversation once
with the Stanford history faculty
and it's one of the few times
in my life concerning metadata
that they truly stunned me, and
I didn't quite know what to say.
So, they were not in
favor of library metadata.
What they said was they're
interested in, they love metadata
and that describes their
collection, but they wanted metadata
that was focused on their
collection, that described it well,
as opposed to having
more generic metadata
that could link across domains.
And they found that
not helpful for them.
So, and since the whole point of
everything that I do is trying
to draw things together through
metadata, I was very puzzled
or surprised to hear that there
was a large group of faculty
which are not necessarily
in favor of it.
So, I certainly wouldn't
say that it's second class,
but you have to realize that it
does have that ulterior motive
and not everybody is
happy with how it works.
>> DAME WENDY HALL:
Go on, Jim, Please.
>> JAMES HENDLER: So, I would throw
in, I think what people miss is
that metadata has a completely
different purpose than the data.
So, it's not that it's second
class it's that it's different.
In the database community where
you're traditionally looking
at what's inside one entity that you
understand very well no matter how
large it is, that notion of metadata
as a second-class citizen
becomes clear,
because we know about our data.
We're using the same store,
we're doing mining, etcetera.
But when you start to move into
I use the acronym DIVE in one
of my talks, DIVE into data.
Discovery, integration,
validation, and exploration.
When you want to move into those
things, you know what's out there
that I can use, how do
I bring it together?
Did the thing that
came out make sense?
Right? So, again we run
some data miner on something
and it's telling us something that
clearly doesn't pass the sniff test,
we've got to watch the watchers.
And then, you know a lot
of the human perception,
you look at the graphs that Katy
was showing, humans are much better
at looking at those than
machines are still in terms
of understanding whether
they're telling us anything new
and different.
And so, those kind of
problems don't live
in the traditional database box.
In fact, I would argue
for archiving and things
like that the metadata is
probably more important.
For example, we're going
to have to do data triage.
Right? There's no way that the
amount of data being produced
on a daily basis, when you take open
data, social media data, web data,
changing data can be preserved.
But a lot of the metadata about it.
So, we could know that
there were articles
about this particular topic being
published at this particular rate
at this particular time, even
if we couldn't keep track,
articles we can probably keep.
But a lot of the videos, a lot
of the details of real time,
when we talk of the internet
of things and sensor data.
You know we're not going to be
able to preserve the sensor data.
We're going to be able to say this
many people were using this kid
of sensor for these
kind of techniques.
And I think, that's to me when the
semantics comes back in as you said.
And that's why so many other parts
of this linking stuff together has
to happen at that stuff
about the data, not the data.
>> LEE RAINIE: Do you know
and the Dewey Decimal
System speaking of metadata.
Dogs and cats, just to go back to
the pinpoint of this conference.
Dogs and cats are archived
as industrial agricultural
equipment [laughter].
>> DAME WENDY HALL: Really?
>> LEE RAINIE: Because that's what
they did in the last 19th century.
They worked on farms.
>> JAMES HENDLER: Yeah, so when
I mentioned the temporal stuff I
should have mentioned it
changed over time [laughter].
>> DAME WENDY HALL: Ah yes,
we should have Wolfgang naming
here for the temporal stuff.
We haven't talked about
temporal stuff.
I did invite him but he didn't come.
Does anyone in the audience want
to pick up a piece about metadata?
Because I actually think
it's different but as
if not more important
in some circumstances.
Do you want to say anything?
No? Oh, okay.
Well you could [laugther].
Hang on, hang on hold that thought
because of the recording
Jefferson that's all.
>> JEFFERSON BAILEY: I do
think there's recognition
in the library archives community
that human provided descriptive,
or technical or administrative
detail will obviously never scale
to the rate at which digital
collections are being acquired
and ingested.
So, I don't know what the
future is, machine learning,
automated extraction,
semantic entities,
it will have to be highly automated.
And anything automated of course
is human created, but there's going
to be all sorts of levels of
mediation that are sort of opaque
to users and I think that's
something to think about.
>> DAME WENDY HALL: This
was the point Philip made
in his talk, this will
have to scale.
This has to be automated.
>> JAMES HENDLER: So, I'm going
to disagree to a slight extent
such as I believe the extraction
and collection will
have to be automated.
But I think if we're not using
human concepts to organize the data,
then we're not going to have humans
able to find the data, right?
So, I think again, one of
the things that I believe
and when I was talking
to Vint before
and he was talking about bottom up.
Well what won't happen is we
won't get a Dewey Decimal system
that designs the whole
world from the top down.
We need something that
can merge bottom up.
But a lot of the bridging, a lot
of that semantics is going to have
to come from a human saying, these
are key concepts in a domain.
Now we ca use computers to
extract to those concepts or type.
So, I don't believe that
just clustering and things
like that is going to get
us where we need to go.
Assuming it's us as humans
who want to do the research.
>> PHILIP E. SCHREUR: So, there's
one other thing I wanted to add
as well, one of the things I
mentioned but we haven't talked
about very much is
that there then needs
to be a commonly agreed
upon vocabulary
for pulling those things together.
And one of the quotes that somebody
mentioned once which I loved,
and I use it whenever I can.
The wonderful thing about ontologies
is that there's so many of them.
So, when I was working at High Wire,
I had extremely the most
bitter fights I ever had were
on vocabulary development
and between groups
like the American Heart
Association, the American Society
of Physiologist talking
exactly the same system,
but they fought bitterly
to have their vocabulary
and their structure represent
the data, so, the metadata.
So, it's one of the big challenges
we have, not only within a country,
but across countries to come up with
that vocabulary that can be used.
>> JAMES HENDLER: But imagine doing
the linking thing you're doing
to data to those vocabularies.
So, Tim Berners-Lee and Ora
Lassila, and I wrote an article
about that 15 years ago
called "The Semantic Web"
and that's really what we were
talking about is that we have
to link the concept space, right?
So, this group can
represent their data,
but someone else can help map it
and then we can take those
mappings and map it together.
>> DAME WENDY HALL: Ramesh
wants to come back here.
I can't see the people
holding the mics,
they're hiding behind the pillars.
Where's the other, ah
can you sort of stand
in my line of sight, so [laugther].
>> RAMESH JAIN: I really
think that metadata is one
of the most misunderstood thing.
And we started in the term context,
now we just heard the term
ontology and languages.
The fact is languages always divide.
Whether the signals are declaring
original information unifies people.
And that's why we got rid of a very
different perspective on metadata.
And the metadata should
not be considered something
which is secondary,
and which is qualifying
and some central piece of data.
When we talk about books,
we consider that books
are the primary thing.
Anything associated with that is
going to be the secondary thing.
But those data are gone.
When you look at a photo
and in the next panel,
possibly we talk about that.
But what is metadata
and what is data?
I think that distinction disappears.
So, what, how is it that we, I would
like to hear your perspective it's
in the days when all this
different types of data is coming
in a distance, how do we start
really considering what makes the
semantics of the data emerge?
It's not one data, another metadata,
it's just different extremes
of data, and there the
semantics is emerging.
>> DAME WENDY HALL: I
was struck when I look
at Katy's visualizations,
you're visualizing, well that,
this is incredibly hard work to
produce those visualizations,
I think because you've got to,
I shouldn't be talking
about this, you should.
But it's there that to me what
about them metadata in that process,
does that make any sense to you,
I'm not explaining it very well.
>> KATY BÖRNER: So, I don't
think you will like this
but there are datasets where
you could have way more metadata
than the dataset itself.
Because if you have
multiple insight needs.
>> DAME WENDY HALL:
He likes that, yeah.
>> KATY BÖRNER: So, every single
insight need will actually require
potentially a reconceptualization
of what's in that data.
And depending on how
you represent that data,
you can run different
types of analysis,
you can get different kinds of maps.
And you get different
types on insights.
I do agree with Jim
that it is very helpful
if you actually have a
few human curated kind
of data dictionaries
as I would call them.
However, I do also
agree that one scales,
so very soon we will have a
very heterogenous landscape,
where you have some human generated
kind of historical data dictionaries
if you envision and the
Dewey Decimal Systems
and then also organizations.
And at Stanford University
you have NbO which is aligning
like 300 plus different
ontologies just
in the biomedical domain itself.
But then you will have also
may data derived structures
and you will want to
interlink those.
You'd want to use the machine
intelligence whenever possible.
But then if you have historically,
or current structures then you want
to interlink it to those
machine generated ones.
And I don't think there
is much work yet on that.
Even though they are now programming
languages which give certain task
to machines and certain
tasks to human beings
on mechanical [inaudible] etcetera
and then it goes back onto tasks
that machines are better
in doing, and certain tasks
that the human users
are better in doing.
Realizing that heterogenous
compiling and running
of code will also be visible
in how we deal with data.
So, I see a very heterogenous
future for that.
>> JAMES HENDLER: Yeah, so in those
four or five hours I'm not going
to spend, let me just say that
the notion of metadata is driven
by the web to smaller
multiple link thinks,
not these very large catalog
driven very heavy weight metadata.
So, the traditional notion
of metadata in the library is
like the Dewey Decimal System.
The thing we need for data
is a small set of terms
at a very high level and then sub
communities that can expand them
in their own ways and provide
linking through this graph
and network of this complex
entity and I think when you look
at that model you can see both
scaling you can see places
where you can tie machines.
So, we've been doing that
with extraction and things.
So, again I think it's where
the technology space is starting
to hit the conceptual
space, but at the moment,
the real problem is no one owns it.
That's what I find the
biggest frustration.
I go to a funding agent and they
all say the same thing to me,
which is that's a great idea
I wish we funded that, right.
And so again, it's very hard.
It's an infrastructure need.
It's a library need,
it's an archiving need
and those are very complicated
at the moment to find ways to do
at the kind of scales we need to do
to both really both
understand the web.
Wendy and I were 10 years
ago co-authors of an article
with some other people
including Berners-Lee guy who,
called "Creating a Science
of the Web" and it's led
to something with all websites.
But you know again,
it's studying that kind
of interdisciplinary must
have social science involved.
It must have library
science involved,
but it also needs the technical
and fundamental engineering skills
for the web, it's a very complex
space, but you know the fact
that you're all here is talking to
the importance of this and the keys.
I promised Wendy I'd
mention websites.
>> DAME WENDY HALL: You
have two or three times now.
Anyone want to take
another direction?
Yes, to my left.
>> SPEAKER 1: Hi my name is Baha
Ecminar [assumed spelling] I work
in the Library of Congress here.
I wanted to ask you the question
about helping us contextualize
a good and effective role
for the Library of Congress
in the century ahead.
So, in the past, we
have had plenty of data,
but most of it publication
related, manuscript related,
traditional library items.
We're beginning to get
large amounts of data in,
but we don't always collect it as
data, we collect it as an item,
or as related to a publication, so
if I can make the question concise,
what can the Library of
Congress do with its consistent
but very finite set of resources
in the century ahead
to make data valuable?
>> DAME WENDY HALL: Yeah, I think
everyone will want to answer
that I will talk to it,
but Katy you go first.
>> KATY BÖRNER: So,
I liked Jim's question
about how do you index data.
And one way that you just mentioned
was to connect it to the publication
in which said dataset was used.
This way you can potentially
even rerun that analysis you get
to these rerunnable papers
which is what many of us want.
Another way is also to connect the
papers now to software artifacts,
to the outsource, which they
are typically connected to,
but now that Orchid is in
existence, try to make people use it
to find carrots for
them to actually use it
and then you get these
very heterogenous networks
where you can pull out one
author and all the artifacts he
or she created; be it
datasets or software artifacts,
or papers, just all come with it.
And I think that will create
much richer search structure
and much richer result sets
when you actually start querying
those datasets.
And it will really leads
to more reproducibility
because you know whom to
ask, how to run that tool,
or that workflow etcetera.
>> PHILIP E. SCHREUR: So,
the one thing I would add
is something the Library
of Congress is doing now.
So, I mentioned there
is this transition
in library metadata
from Marc to link data.
And the one thing that Elsie has
really led us on is this development
of a new schema called
BIBFRAME in RDF, to be able,
as something which will better
represent the library data.
There are a number of schemas out
there, but they really took the lead
in trying to develop this schema
to represent library data.
It eventually will be community
driven, but they really pushed it
to get it going and
that was a great effort.
>> DAME WENDY HALL: Well, yes.
>> JAMES HENDLER: So, the only thing
I would throw in is I think the,
so the standards that everyone wants
grow from defacto standards of use.
And I think this is a place that
has the authority, the power
and the driving need in looking
at archiving the web and things
like that to be the place that
some of that work could be done.
And you know, again, I don't know
enough about the organization,
the internal structures, but I
think things that grow out of here,
and that look at those technologies,
look at the vocabularies,
look at the new things, say,
here's stuff that can be done,
would be something that would
attract a lot of attention and grow.
>> DAME WENDY HALL: Can I just
add that I think librarians,
there's always been a lot of
data science in libraries.
I think there's a big
shift for the next century
where if I were the Library of
Congress, I'd be investing a lot
in data science training
for my staff.
And I don't just mean from the
concept of library data science,
I mean just generally getting to
grips with data science and get
as many people as possible in the
library understanding this world.
Because in the future, libraries
will be data warehouses, largely.
And that, I was running the
place that's what I'd do.
But I'm not, so.
Is that a hand up there?
Yes.
>> LIZ MADDEN: I'm Liz Madden
from the Library of Congress
and this is more of a plea from
the bottom up to all of you
who have students in
colleges and things.
I work with the data, the
metadata in particular
and also the regular
data that comes in.
And a lot of the interns and things
and the new Library students.
Along those lines it would be really
helpful for people being taught also
to understand basic concepts
like relational databases.
Because I think that's what
we see when we get datasets.
I looked at the map of the domains
and I saw how tiny the US was
and I know that's just
because they use .US
and that's a database error to me.
And so, that kind of concept seems
really basic probably to the people
in this room, but it's a really
foreign concept to a lot of people
who are studying humanities
and stuff like that.
So.
>> DAME WENDY HALL:
That's very interesting.
I started teaching relational
databases when I first taught
as a computer scientist in 1980s
I don't know whenever it was.
It's interesting you
should bring it up now.
I would widen it beyond relational.
I think it's about
understanding the world of data.
No, actually.
No, I would disagree with that.
I would disagree with that.
Sorry, Jim you teach databases.
>> JAMES HENDLER: Yeah, I would.
Well I don't teach databases,
because of computer science
relational has always been
with databases, what
I teach is about data.
And I think understanding that
difference that databases are one
of many ways of holding,
thinking about using,
retrieving, storing, etcetera data.
That's really the key.
And you know again my early work
in AI led to sort of semantic data,
but then through realization
that it's
about describing all
that other stuff.
So even if you're, so the
problem is once you're
into the relational database, you're
forgetting all that outside context.
But, that said, I think
teaching students about data.
We, at RPI, I'll put in
a plug are actually going
to start requiring every student.
So, not just the computer
scientists,
but every student is
going to have to take data
in tons of courses, right.
So, we think data is to this
generation what writing was
to the last one, the
thing they didn't learn
in high school that
they should have.
And we're going to
really push hard on that
across the entire curriculum,
across the entire university.
>> DAME WENDY HALL: And
digital literacy generally
for the entire population.
Oh, one over, I think
this is probably going
to be our last question so we have a
chance to hand over between panels.
So, yeah. Tell us who you are.
>> SPEAKER 2: [Inaudible]
and I'm a student
from the George Washington
University.
So, I have a question
about as a way to enter
into the largest scale datathon,
people may face an issue how
to solve the storage problem of
the being or treating and data,
especially for the metadata
because the already size is large.
So, when people try
to recover them how
to solve these problems, thank you.
>> DAME WENDY HALL: Thank you.
Jim, you're nodding.
>> JAMES HENDLER: Yeah,
it's a great question.
In fact, when I was
first asked to be
on this panel I thought that's
what I was going to talk
about until I realized I really had
nothing very smart to say about it.
But it's clearly the challenge.
And the scale with which our ability
to store data has been growing,
has up until recently been keeping
speed with the increase in data.
That, you can look at almost any
company that's publishing those data
now showing we're heading
for the crossover point.
So, one of the hardest things
I think we're going to have
to learn is how do
we throw away data.
How do we decide what to
collect, what not to collect.
And I don't mean now in the sense
of meaningfulness in the sense
that an archivist thinks about
whether we're going to store these
or those, but really in
the large hadron collider,
it generates so much data that a lot
of what the different
sensors do is figure
out which data they can ignore.
Right? so 90% of what gets
created never gets put
into the petabyte that
they generate.
And it happens at a very low,
automated level and
things like that.
I think we're going to have to
start really grappling with some
of those kinds of issues.
So, not just what do we want in
the library, but must we keep
because there's going
to be a lot we can't.
And I think one of the
reasons people tend
to move toward journal articles,
to books, and even to Twitter is
because those things tend
to be bounded in such a way
that we can think about
keeping all or almost all of it.
But when you start moving to you
know every video that was take
on the web yesterday, right
on anybody's phone yesterday,
every Vine, every sorry I'm going
to get on my favorite soapbox,
I'll get off of it, but again,
we're going to have to figure out.
I mean a lot of that is
crucial, but a lot of it isn't
and in some way those kind
of collection management type
issues are going to be very,
very crucial to the next
generation of what you call do.
>> DAME WENDY HALL: I don't
think storage is the issue,
personally I think.
I was just looking up
something my colleagues
in the Optical Research Center
are doing in Southampton,
where they're predicting,
you too yeah?
They're predicting, you know they've
got a huge store as much data
as you can possible want.
And so, I think material science is
just, is going to give us solutions
to the, I think the problem is
making the haystack bigger doesn't
enable you to find the needle.
I mean this is the problem.
I think discovery and navigation
is going to be the problem.
And the question is do
you store everything,
or what do you decide you know to
throw away, if you're not storing,
I mean Ted Nelson who started the
whole hypertext thing always argued
you absolutely store everything.
That's always been his position
and he's never changed it.
And I think that's a
big issue for debate.
And it's on that the
libraries have got to tackle.
You know they are the first
frontline of this really.
Do you want any final comments
from our fellow panelists?
People feel, no?
Okay. So, it's a bit, everyone, I
can sense the room is a bit hot,
so we're going to do a
handover, there's water.
There's coffee.
I'd love there to be tea, I
don't know if there's tea.
It's 3 o'clock, I'm a Brit.
And there will, we'll just unmic
this panel and mic the next panel
in just a five-minute break.
And I'd like to say
thank you very much
for the panelists this afternoon.
[ Applause ]
This is the final panel
this afternoon.
And it's chaired, I'll let him make
the introduction to the panelists.
It's chaired by Matt Weber who
it's been my great privilege
to work with.
We have been working on this event
for a year I think, at least.
And Matt with Jimmy and
Ian organized the datathon,
but also he helped me with David and
Nasha Contractor [assumed spelling]
who unfortunately couldn't
be with us today
because he had to be
in France, anyway.
No, he has a very genuine reason
for not being able to be here.
But the four of us put
this symposium together.
But it was really Matt
who did all the work.
Well, together we did it, but Matt
was the person who was there for me
when we were organizing
this and working
with the Kluge Center
to get this to happen.
So, Matt, now it's your turn to
chair this panel over to you.
>> MATTHEW WEBER: Well, thank
you to Wendy for her leadership
in this it's really a privilege to
be here and to have an opportunity
to chair this closing panel.
I just want to make a few
quick remarks and then get
to the meat of the conversation.
The title of this panel is
"Saving Media" and I always think
of news media because that's really
what I study day in, day out.
And on that topic, I was looking
at Facebook last night and happened
to see a post by one of my
graduate students who was talking
about her 7-year-old daughter
and she was reporting a story
of a conversation she'd had
with her daughter last night.
And her daughter had said
to her, 'did you know years
and years ago people use to get
their news from papers' [laughter].
And she's reporting the
story back and she said,
'I turned to my daughter and I said
yes, they're called newspapers.'
And her daughter looked at her
and said, 'were you alive then?'
[ Laugther ]
And that's what it's come to folks.
But this idea, not that we
necessarily need to have newspapers,
I'm very split on the idea of
whether we'll have printed papers
in the next 40, 50 years.
But that is not the
goal of this panel.
But this idea that we
need to preserve the media
and preserve the content that we're
producing today is front and center
to the discussion that
I'm hoping we're going
to engage with right now.
In my own research, I
spent a lot of time looking
at archived digital news media.
And I found that it's
critically important to have rich
and detailed archives of news
and of events that have occurred
across the globe for many,
many different reasons
and I'll give you just
two really quick examples.
One, is go back to for instance
the events that occurred
around Hurricane Katrina.
CNN at the time was online news
source that did quite a bit
to cover Hurricane Katrina.
And created their own special
archive for Hurricane Katrina
and you can go to CNN's website
and I believe it's still on there.
Did not check just
now to see if it is.
But they have a repository of
Hurricane Katrina news coverage.
And if you go and you
look at that, it's been,
it's essentially been cleaned up.
It is not the coverage
that you will find if you
for instance go to
the internet archive.
And if you go to the
internet archive and look
at the very same coverage from
CNN and actually look at what was
on the web at the time,
it's a very different story.
And the point being, we need
accurate archives because we do need
to understand actually
what occurred at the time.
So, as we start to think about
these issues and importance
of archiving news media in
the modern landscape of media
and social media, just a couple
thoughts that I'll put out there
as you hear our presentations today.
Think about this idea of what
is news, and what are media.
What are we saving?
What is a journalist?
Who is reporting on an event?
Who is reporting on news?
And how do we record that?
These are clearly hard questions
to answer, but these are some
of the challenges that we face in
thinking about saving the media.
So, with that I'm going to
introduce our first panelist
and then I'm going to take a
seat down here so that I can help
to keep us on track today.
And our first panelist is
my colleague Phil Napoli
from Rutgers University.
Phil is a professor in
journalism in media studies as well
as the associate dean of
research at the school
of communication and information.
Beyond being a fantastic colleague
and someone I have been
fortunate enough to work
with for the past couple years, Phil
is leading a new project looking
at local news media and
developing new methods and measures
for understanding local
news media coverage.
And then, lastly, I will say
that unfortunately Phil is
leaving Rutgers University
and breaking my research heart.
And joining Duke University
as a James R. Shipley
Chair of Public Policy.
He'll be starting there in the fall.
So, with that I'll hand
things over to Phil.
>> PHILIP NAPOLI: All right thanks.
So, seated I'm good?
>> MATTHEW WEBER: What's
your preference?
>> PHILIP NAPOLI: I'm happy to sit.
I don't have any slides
or anything like that.
Hi everybody.
Thanks, Matt, I'm talking today I
guess really as a social scientist,
as a wannabe user of all of this
archival data that don't yet exist
to serve all of my needs.
And as Matt mentioned I sort of
have interest in the relationship
between media and public policy.
And so, I wanted to
kick off really sort
of continuing what you were talking
about which is why
this matters so much.
Just to give you an example.
I came here because you know being
here in DC reminded me of something
that happens in this
city every four years.
And I hope this will sort
of illustrate why this is
such an important area to archive,
but every four years the Federal
Communications Commission is
required by Congress to
evaluate the adequacy
of its media ownership regulations.
And you're like what's he talking
about why does this matter?
But this is something I get
involved with every four years
and for you know quite a while now.
And the fascinating part about it is
that every four years the FCC
commissions a series of studies
and the goal of these studies
is to determine whether
or not these regulations
need to be altered,
strengthened relaxed etcetera.
And every four years most of these
studies say the exact same thing,
they all have a disclaimer that says
something along the lines of yes,
it would have been nice to
really investigate this question
in a rigorous way, however,
the data don't exist.
So, you know the FCC is trying
to answer questions like,
you know what effect
does concentration
of media ownership have on
the diversity of the quality
or the quantity of local news.
Is there some relationship between
the characteristics of the owners
of various news outlets
and the ideological
orientation of the content?
And literally every four years they
try to remake the wheel and try
to come up with studies and
they've never developed any kind
of systematic data archive
to answer these questions.
And every four years the
studies basically say,
yep we're a bad study,
but here it is and go try
to make decisions based on it.
And to me it just highlights
how important having that sort
of archive of the performance
of our journalistic system,
and especially there was a point
I really want to emphasize today,
is that this sort of archiving,
and we haven't talked about it much
in this dimension has to
drill down to the local level.
And that's something
that we're dealing
with now in our current research.
So, you know from a lot of this
type of, these types of concerns,
the FCC did something else recently,
they wanted to assess the extent
to which community's critical
information needs were being met.
I don't know if anyone followed
this, this was a couple years ago,
and just doing that research was
so politically you know toxic,
that the research got cancelled
because they were actually going
to go in and actually
talk to journalists
about how they do their job.
But nobody had a problem,
which is kind of interesting,
with the process, which has
been happening for years,
of actually studying
journalistic output.
But once again the data aren't
really there to do that.
So, we've been working on a project
where, we're essentially going
to gather, with the help of
the Internet Archive and Matt,
the totality of journalistic output
for a sample of 100 communities
in the US, for only about 7
days, I should mention that.
And the idea is whether they
are a television station
or a radio station, or
a hyper local news site
that you're web presence represents,
you know some reasonable indicator
of your commitment to
produce in journalism.
And the idea is to start to try
to inform this other very
pressing policy issue today,
or course which is
essentially what do we do
about the ongoing crisis
in journalism.
And how do we inform
efforts to try to address it.
So, there are a number of
foundations, fortunately,
that we have here in the US that
really want to provide assistance
to keep local journalism robust
in the face of the variety
of technological and economic
conditions that are making it a less
and less viable enterprise.
But they don't know.
You know what are the real
contours of this crisis.
What kind of communities is
the crisis worse versus better?
What are the conditions
that help us understand
where this crisis is
really pronounced?
So, we're trying to gather that kind
of data essentially on the output
and the performance
and the infrastructure
of journalism at the local level.
And it was interesting as we started
to do this work we realized there
really is no precedent for it.
You know the archiving
of journalistic content,
especially at the local level.
If you wanted to do a study on
you know local radio's coverage
of an issue from six months ago,
you'd have basically nothing.
Local TV same thing.
And again, we know this is the case
when we talked about online content.
So, this entire media ecosystem,
and we use that term a lot
in our research now to really try
to understand the interconnections
between old and new
media, but also recognizing
that our media have converged
to a sufficient extent now
that you really can use web content
as a window into the totality
of a community's media ecosystem.
And so, that's where the archiving
of this content becomes
so important.
Because we don't know, you know we
have these broad generalizations
about what's happening in
journalism, and in fact
but very little of it is
based on actual efforts
to understand the context in which
journalism maybe matters the most,
which is that local level.
The community media.
The context in which
your vote etcetera.
You know when you look at
these issues from a national
or international level, the same
kinds of problems don't exist.
There are really great archives
of nightly network newscasts
and things like that.
But if you wanted to get a sense
of how the local news was covering
something like Katrina, or whether
or not you know different types of
levels of competition had positive
or negative effects on
the state of local news.
Answering questions that could
truly guide policies directed
at making sure that we
have a media ecosystem
that really serves citizens well,
that creates an informed citizenry
that really, you know enhances
the democratic process.
We have been flying blind in
that area for a very long time.
And again, that's the irony is
that it's not a web based problem,
it's something that's characterized,
all of our electronic media leading
up to this that you know, it's
still easier to do a study
of news coverage from literally
1940 than it is to do a study
of news coverage from
you know a year ago.
Because you know what they did a
great job of archiving newspapers.
It's all there, but if you want
to you know really
study the complexities
of today's media ecosystem,
it's challenging.
And so, that's what's fun
about what we're doing
with the internet archiving.
And Jefferson is fortunately saddled
with solving this problem,
even there.
We handed them a list of URL for
all these local media outlets.
And they said, jeez, yeah, know
this falls below our radar a lot
of this stuff, because
it comes and goes too.
Local media these days is so,
you know outlets come, they go,
you know they go out of business.
It's a very you know
sort of dynamic space.
And it's a space, that which
again the audiences for a lot
of this stuff fall below thresholds
that you might think would matter
for a sort of large
archiving project.
But we're trying to drill down to
this local level now, and hope,
you know, trying with the logic
that whether it's the economy,
or the environment, or jobs,
or education that we've always
had these regular indicators,
these ways we could take the
temperature of how's the economy
in this state, or how's
the economy in this region.
We've never had that about, you know
hey how's journalism doing here?
Because journalism whether,
you know all the ways you learn
about the economy,
education, etcetera journalism,
at least traditionally has been,
maybe it's decreasing extent,
the mechanism for that.
But we have no, we have sort of the
you know thank God for you guys,
you know the state of the news media
that we have sort of broadly right?
But now, you know can
we drill down deeper?
And sort of understand
how a community
in New Jersey's journalism,
you know ecosystem has evolved?
Is it getting stronger?
Is it getting weaker?
And does it need help?
So, these are the big questions
we're trying to deal with
and thank God folks
like Matt and Jefferson
and the Internet Archive are
helping develop the tools
that will allow us to
do this kind of work.
But again, it's not just you
know historical internet,
it's truly trying to inform policy
that you know that it's that kind
of relevance to this kind of work.
So, I'll stop there.
Thanks.
[ Applause ]
>> MATTHEW WEBER: All
right thank you Phil.
Our second panelist is Ramesh Jain,
he's the Donald Bren Professor
in information and computer sciences
at University of California Irvine.
We were fortunate earlier today
to be joined by one of the fathers
of the internet, and we are
now joined by the father
of multimedia [laughter].
>> RAMESH JAIN: So,
everybody's awake?
Looks like that.
So, what I'm going to do is to
talk about some of the issues
that in my opinion are
burning issues particularly
when we are talking
about saving the web
and the way information is
nowadays getting created.
Let me start with one simple chart,
and that is in fact if we think
about it, this century
is really very different
from the last century.
There is a pretty interesting
diagram
that maybe Miguel put last year,
or somebody else put it last year.
When you look at this diagram,
this brings in some very
interesting perspective.
When we start looking at
this diagram what you find is
that these are the analog
cameras that were being sold.
And then you suddenly come here
and the world completely changes.
And when you look at these trends,
as you see this trend really started
taking place between 1999 and 2001.
This is very interesting
and serious implications.
Matt told you a story about the
daughter and about then newspaper.
I had a very similar experience,
not secondhand but firsthand.
My own granddaughter who
is now about 8 years old.
Two years ago, she comes to me with
a CD and says she calls me nana,
she says, 'nana my mother
is not telling me the truth.
She says that music comes
out of these CDs,' right?
'Out of this device.'
And I said, 'yeah that it true.'
She said, 'no that is not true.
Music comes out of iPod.'
And that's part of the world
happening, but more interesting
that that is really
this particular case.
When you start looking at the
world, for most of us here,
and this is very closely related
to what we were talking about data,
metadata and all these
kinds of things.
For most of us, until very
recently we tolerated photos,
we tolerated videos
in our documents.
We still have to pay
extra for color photos
for getting things printed
and things like that.
That was the world
that we grew up in.
That's the world that
people are thinking even
in this room indirectly or
directly about libraries.
What is happening is that the
world has changed completely.
When you start looking at
the world, the type of thing
that people started doing, the type
of devices that started coming,
the way information started
getting created is very different
for young generation now, at
our age, we tolerated photos
and videos in test, reluctantly.
For the young generation, they
go to Instagram and Snap Chat.
Why? They don't want to use text.
They tolerate text very reluctantly.
And that's the world that
we now have to consider.
That's what is happening.
So, when you start considering this,
it has very serious philosophical
and long term implications.
When you think about the world,
and when we present the world,
there are two things that
we use, objects and events.
Our current thinking, and
dominated by computer scientists,
we deal with the object-oriented
programming,
we talked about object-oriented
databases
and all these kind of things.
We were very object-centric.
But in the modern world when we are
dealing with all of these strings
of data coming in etcetera.
And when we are dealing
with Twitters of the world,
when we are dealing with Facebook,
events are becoming as important,
possibly more important
than object space.
What does that mean?
That means the following thing.
We have to learn how to represent
events and how to deal with events.
And when we deal with
events, these famous 5 W
and 1 H become extremely important.
For events, when you want
to represent events there are
these six important filters.
It started from where,
when, what, why, how?
And all this information
none of them is metadata.
Everything is data.
The pictures and the documents they
are the experiential information.
The informational component is
in order to structuralize that,
the information that
you start capturing,
you start putting in information.
You cannot interpret photos.
You cannot interpret videos.
So, what do you do?
You analyze those,
you use deep learning,
or you use context learning or
whatever to get some information
that you can use for
indexing purposes
and you start dealing
with that okay.
So, that's what you start doing.
So, when you start looking at these
charts, you start seeing that all
that information the 5 Ws
and 1 H it starts coming
in directly from different sources.
And one of your goals becomes how
do we populate on this information.
Okay. Once you do this for event,
then you have some
structure of events.
Okay. So, you can start representing
all those things completely.
And once you have done this then
you can very easily start dealing
with this.
And in computing term you start
representing this using media this
on and whatever way you want to
do and you start populating this.
So, for each event you start
creating this kind of a structure.
Indeed, the structures are good.
You have the substructures,
you have the super structures,
everything gets linked.
Effectively what you start doing
now is, in place of creating a web
of document that you have
created, now you are starting
to create a web of events.
You remember we talked about doing
these things very differently.
When you start thinking about
this and when you start thinking
about doing search, the
way you start searching,
the way you start navigating
also becomes very different.
See, for example, suppose you
are looking at some pictures.
We are told how do we
search for a picture.
That's where you will start
thinking more about metadata prints.
If I want to search from a picture,
I have to go to a steering
wheel of sorts.
We have to get out of that
narrow [inaudible] box.
We have to get out of that and
we have to have the complete view
of this space and we
have to start navigating.
So, when you start navigating
this, in the current world then
for example, in order to demonstrate
this, there are four components.
So, if you start looking for I
want to see all the events related
to this particular picture, or what
happened in this particular thing,
oops I'm going the wrong direction.
Okay, it will show me
all those pictures.
If I want to see at
the same location,
I want to see the same
images that it resulted,
or the same person
who is in the picture.
So, this is how we start
navigating these things.
We are organizing this now
as a different kind of web.
When we start considering this
and we are about journalism
and all this kind of thing,
what Twitter does is
it creates a microblog.
What you require really
is a micro report.
In journalism, the first principles
are you should be objective,
you should try to report as
honestly as possible things.
Opinions should be
separated from facts.
That's the first principle.
The difference between
facts and opinions.
If you want to capture reports,
then in each report for the event,
you have to learn how to do that.
Luckily for us, we can
do that very easily now,
because we have these smartphones.
When I take a picture, my smartphone
is really taking a report.
And in this report, I have
the location, I have the time,
based on this I can do the
contextual reasoning I know
and I can show you a
system working on my phone
where I'm taking each
picture it will tell
that this is the event
going on here.
It becomes a report here,
it will tell me exactly
which room the particular
thing is going on.
It could even tell me that
today's program was available,
who is speaking, and what's
going on in that thing.
So, your report will be
very, very objective.
Of course, you can add
your subjective information
to those things also there.
So, what we did was to see
an experiment in library.
Definitely Flickr from
Yahoo announce
that they will make 100
million photos available
to people for experiment data.
And these 100 million photos
have some of the contextual data,
the cloud metadata, they have some
of the concept [inaudible]
applied and that will be done.
So, what we did was to run
an experiment we said suppose
that in the library of
future all these kinds
of reports are available, these
kind of things, can we go and look
at each photo as a report
or as a micro report
and can get some interesting
information out of that.
That experiment we
did learn something.
And I just want to mention that this
is currently doing for Flickr data,
there's no reason that we
can go and use every picture
that you are taking in this room
or anywhere else can
become part of that.
So, when you start doing
this something really start
happening there.
Suppose that you want to
find out Olympic games
and I must say this example was
motivated because I was invited
by London people here and so on.
So how do we detect Olympic
games from all these photos?
So, what you can do is that Olympic
games take place every fourth year.
And for this I define
some particular concepts.
Now, in each picture
I have the concepts.
So, what I start doing is I
got and plot those concepts.
And this is showing now you
the plots of different things.
I look at the peaks which are
appearing and I immediately find
out that in London there
are Olympic games going
on at this particular time.
And you can see the peaks there that
immediately start showing you there.
You see a secondary peak, this
was for the [inaudible] Olympics
that were going on immediately
after the Olympics and so on.
So, without doing anything,
just from the photos you can just
start getting all this information.
So, when you start thinking about
reports, when you start thinking
about in future what kind
of information we are
going to be archiving.
Everybody talks about every
fourth year, the election,
every this game is
going on and that event.
That's what really happens.
For each event you generate
enough experiential data.
You generate enough
reports related to that.
And later on you the
make comments about that,
and all those things
start getting linked.
So, the important thing is if you
do not do event based indexing
mechanisms, we will be
creating lot of write only data.
Thank you very much.
[ Applause ]
>> MATTHEW WEBER: And
our last speaker
on this panel is Katrin Weller
who is an information scientist
and senior researcher at
GESIS Leibniz Institute
for social sciences
in Cologne, Germany.
She is also one of the
inaugural digital fellows
at the Kluge Center here
and was here last year.
And focuses on the use of social
media as a potential resource
for future historical
perspectives from the viewpoint
of computational social science.
>> KATRIN WELLER: Yes, great.
Okay. It's such a pleasure
to be here today
and I'm already quite overwhelmed
by all the input I got throughout
the day and try to relate to some
of that during my talk now.
I'm an information scientist.
I did my PhD in information science,
but I also studied history
in my magister degree.
So, I also have some experience
in working as a historian
and doing some archival work and
somehow the internet distracted me
and I got into information science.
And I'm so happy that I now can
bring these two perspectives
together again.
What I want to highlight
today is the importance
of also archiving information
about the context of digital media,
of web data, of online data, in
addition to the technical parts
that we may be already
archiving in some project
which we have seen today.
I'm saying that because
there's quite some resemblance
of social media data
and prior sources
that historians used to work with.
But there's also some
editions that are new to that
and that we will keep
track of in order
to make sense of that later on.
So, I'm not so much worried
about getting every single piece
of information, collecting every
single tweet archiving every single
user interaction that
has been on the web.
I'm much more worried about if
we lose the important context
about how these types
of media were used.
How users were actually interacting
through these type of portals.
I mean if children today don't
believe that a CD can play music,
of that news can come in form of
paper, how's anyone to believe
that we use the phone to
take photo and show it
to someone in like 100 years or so.
So, we have to capture
these kind of context.
We have to show people how this type
of information was
produced and is being used.
Historical sources and historical
source criticism as the way
that historians work and
interact with their material,
often deals with text but not,
like text is not equal text.
So, there's a lot of
different types of text.
Like we've got diaries.
We've got newspapers, we've got
official documents like kind
of notes to the people by a
specific ruler for example.
We also got multimedia content.
So, in my magister degree thesis
I was studying film propaganda,
British film propaganda
during the Second World War,
and so I was watching these
propaganda movies, but I also had
to study like the circumstances,
under which they were produced.
So, I had to read a lot about
different types of directors,
about how the funding would
work to produce such a movie,
how the cinema systems worked, who
could afford buying a cinema ticket
to see one of these movies.
A lot of things like that.
There's artifacts of
different sorts.
And, some of these things
directly relate to digital media.
So, for example we've got war
diaries in forms of blogs.
So people blog about their
experience during war.
We've got like, what used to be
the kind of leaflets displayed
to the people can now
become in form of a Tweet
by the president for example.
And we've also got lots and lots
of eyewitness accounts
in multimedia formats.
For example Flickr on YouTube and
on other channels, which are added.
We've got lots of maps,
we've already seen
pretty examples of maps.
And we've got maps that track how
users interact, how they behave,
how they move through space.
So, lots of new kinds
of data in this regard.
This is the comment line
from an online newspaper.
So, we've got news, quite similar
to the news we got on paper.
Or we've got users directly
sharing it in different platforms.
Commenting on it on
the news website.
And finally, we've got people
reporting about news events.
That is the history of the
Wikipedia article about the landing
of the plane on the Hudson River,
and event that was
prominently tweeted
about first before it
entered the traditional news.
And kind of people
also responded to that
by setting up a Wikipedia article.
We can see the history
of how that article grew.
And there's differences
whether you look it
up in the English Wikipedia version
or another language
version of course.
So, lots of different
types of sources.
Okay, so in traditional
historical studies.
For example, if you want
to make sense of something
like that you will have to
know quite a bit about how
to handle such types of sources.
So, numismatics as the kind of study
of coins, is a field where you have
to learn about what does a symbol
mean that is displayed on a coin.
What does it mean if a head is
displayed in a specific way?
What are the symbols that
may be placed around them?
Why are they using
specific types of images?
Why are they writing
specific types of letters?
There's an entire, you can
kind of spend two semesters
in university just learning
about how this works.
Also, how the material was chosen,
how the coins were produced.
All these kind of things.
What if our artifact
looks like that?
What do we know about how
something like that was produced?
What symbols are used here?
Why is something displayed
in the way it is?
What does at POTUS
means in this sense?
We have to know that maybe
this account is moving
on from one president to the other.
Maybe it is staying
with one president.
We have to know how
multimedia is displayed
in this kind of information.
What it means to retweet something.
All this kind of information.
What if we don't get
this, but only that.
So, what if instead
of the visual display
of a tweet the only thing that's
being stored is its tweet ID,
or maybe the JSN file with the
textual information behind it
but not the visual
information around it.
Does that ring a bell with someone?
What's this hashtag January 25?
Well you may probably
rather recognize this one.
So, during the Arab Spring movement
people were quite frequently using
the hashtag Egypt, but only
after like five or six days
after the events unfolded.
Prior to that they were
using other hashtags
like the January 25 hashtag.
So the context we have to be aware
of is these standards may change.
If you want to study an event
like the Arab Spring movement
in the future you will have
to be aware how people
actually talked about it.
What were the key hashtags probably,
or the key references you needed
in order to take part
of the discussion.
The same is for user account names.
So, user account names are changed,
for example on Twitter
are changing a lot.
I only studied soccer clubs
in different countries
and already had trouble with these
soccer clubs always changing their
Twitter handles and really
being in trouble in figuring
out which was the current official
handle of that soccer club.
So if you have the same with
eyewitnesses during events
like that, or with presidents,
or with other actors,
that's getting really complicated.
So, you need this kind of context.
And of course there's lots
and lots of other events.
And this kind of multiplies.
We also need context in order
to understand the evolution
of platforms.
And I think that's kind
of a thing to start now
to preserve the evolution
of different platforms,
how the emerged.
This is one of the first
drawing of Twitter,
how it could look like
as a platform.
How many of you remember how Twitter
looked like in 2007 for example,
or how it changed its
look and feel in 2010.
It's really difficult to understand
that from our perspective today.
So, some things we need to
think about today is how can,
maybe today retrace how
platforms like Twitter evolved.
One thing you could
look at for example,
these kind of things we have books
telling people how to use Twitter
which will give you a lot
of screenshots and images,
which can trace back sort of the
history of Twitter over time.
You've also got videos.
People sharing YouTube
videos of how to use Twitter,
which gives you an idea
of the look and feel.
But it's something you
really need to look into now,
in order to understand how it
works and not leave it to the kind
of people who don't recognize CDs
and newspapers anymore in 50 years.
The same about users.
So, if you heard about, so Lee
told about the demographics
of different platforms, and that's
something you really also have
to keep in mind.
Who's using these platforms
at different points in time?
How are these platforms being used?
How many people are actually people?
Or maybe the ISS Space Station.
And again, what is being
promoted to the public in terms
of how you use these things, so
that's even "Facebook and Twitter
for Seniors, and Dummies."
So, it's kind of quite interesting.
And with that I'd like to close here
and look forward to your questions.
[ Applause ]
>> MATTHEW WEBER: So, we have plenty
of time for questions I have a list
that could probably
take a half hour.
I'm going to take moderator's
privilege and ask just one question
and this actually isn't my question,
this is a question that was emailed
to me by Edward McCain, who works
on digital journalism preservation
at the University of Missouri
and couldn't be here today,
but he said I have a burning
question I have to ask.
And it's a great one.
Which is he said to me over
email, what are the ethics
of news collection, and we could
say more broadly with this panel,
collecting media around events,
collecting news around events.
What are the ethics of news
collection without permission?
Because in none of these cases
were we talking explicitly
about asking people can we
take your local news site,
can we take your photos of
yourself at Sydney Harbor.
So, what are the ethics of
collecting without permission
and should we or shouldn't
we be doing that?
>> PHILIP NAPOLI: I, you know
thinking strictly in terms
of ethics, to me there's
the thornier question goes
to the issues of legality.
Because I feel like ethically once
something has been disseminated,
the act of keeping and analyzing
should almost be implicitly
permitted by the act
of disseminating.
But, as long as I can keep going,
you know it's not with the goal
of further commercializing it right?
Again, as someone who's worked
in this space for so long,
it's been so frustrating to
watch as sort of research
with a you know legitimate public
interest purpose can be stymied
by denials of access to
content that's already been made
publicly available.
That's, so to me it's you know it's
more these issues that have come
up in the past about issues of
copyright and things like that
that I actually don't know
the answer t. I just know
that to me there's the yeah, I
mean from a research standpoint,
those seem to be the
greater challenges,
but you know maybe that's just me
being a totally heartless social
scientist there, but
I just, I mean ethics,
I mean you know again you
know naivety, I don't know
but I don't think ethically given
that it's already been disseminated,
that if someone wants
to keep it and study it.
>> MATTHEW WEBER: What about then,
the photos of myself at an event
or in from of the Sydney Harbor?
>> RAMESH JAIN: Right, I think
it's related two things here.
Number one is really what the
first issue you are talking
about is the privacy issue.
Whenever I'm asking somebody whether
I can take your picture or not,
whether I can do that or not.
That's more of a privacy issue.
And when you are in public
spaces, usually it is considered
that that's a weaker issue.
Because people are going to
report in the public spaces.
So, but that's a privacy issue.
To me, ethic issue is more
in terms or reporting.
Are you reporting objectively,
or are you completely
reporting subjectively.
Now, what is very interesting
is that even using pictures,
which are considered
normally objective,
people have used the picture
without giving you the context.
And then they start
talking about the picture
as if something else had happened.
And there are lots of things
that you can find on the web,
where they will show that whole
pictures misrepresent the situation.
Unless you have all the
information related to picture,
which is nowadays very
easily available.
So, when you take picture
from a smartphone.
All the things about location,
time and the in which
context you took the picture,
etcetera is usually available.
That starts making it a
little bit more objective.
So, to me I think the issue
in reporting particularly
with the journalism and particularly
for events is more am I being able
to differentiate between fact and
fiction and am I facts and opinions
and am I representing those
things very correctly or not?
>> MATTHEW WEBER: Which
is not easily done.
>> RAMESH JAIN: Which
his not easily done.
>> KATRIN WELLER: So, back home
with my colleague Katarina
Cinda-Colanda [assumed spelling].
We are doing a project where
we're interviewing social media
researchers, or people
from different disciplines
who study social media for
different types of purposes.
And within these qualitative
interviews amongst others we've
asked them about research ethics,
what kind of ethical
considerations they have
when working with social media data.
And that's been quite
interesting, because they're really,
really different perspective, often
it drills down to privacy questions.
But you've got people saying,
well it's all public anyway
so we don't really have
to care about that.
But we also have these kind
of people say that's really
a very complicated topic.
Lots of people may not be aware,
or may not have been aware
of how public things get once
they, for example post a comment
on the public Facebook page,
and they still feel like being
in a protected space where only
their friends can see what they
are doing.
And of course, it aggregates,
once you work with aggregated data
and kind of run some sort
of algorithms on them,
who will for example
then say something
like this person's likely voting
for their party, or likely kind
of being a terrorist, or whatever
and this implies new
forms of things.
We also got people who kind of feel
the obligation to share the data
because some people want to be read.
So, we also have this affect that
once something got really viral
on the web already, or something can
be attributed to a specific author,
that you get a kind of feeling
that you have to quote this person.
So, it's really an untrivial task
and really something that's
still being very much debated
in the community of researchers
who work with social media data.
>> RAMESH JAIN: I just want to
add one more thing to what I said,
this [inaudible] item that some
particular person doesn't want
in his events, report us to come
and report objectively, okay.
Now, is it an ethical issue?
Or what is that?
Because these reporters
are trying to do their job,
they are not being allowed to come
because somebody's
privacy is being violated.
Privacy or the perception that
that person wants to prove it.
>> MATTHEW WEBER: We're
treading dangerously
into political territory.
>> RAMESH JAIN: That's why
I'm not mentioning any names.
>> MATTHEW WEBER: Without
mentioning names.
And with that, thank
you for the comments.
>> RAMESH JAIN: But that's
what the question is.
>> MATTHEW WEBER: No, I
think it is important.
>> RAMESH JAIN: That is
exactly related to that.
>> MATTHEW WEBER: Well
and then can others step
in as reporters and
objectively report?
I think with that we
should open it to questions.
We've got a couple over here.
>> SPEAKER 3: Clifford
Stultz [assumed spelling]
who is a political scientist
in the UK did some early work
on analyzing the causes of the
London riots a couple of years ago
and he was early, and
he used YouTube reports,
contemporary YouTube
reports that people involved
in the riots have posted
up onto YouTube.
But he commented on the fact that
he would use these and then a couple
of days later they would be
taken down because the people
in question had learned that
the police were using YouTube
as a source of information
to try to identify people.
So, perhaps I'd just like to
ask you to reflect on that
because you talked about privacy,
you talked about copyright.
And the fact that these
activities are not morally neutral,
either that we are
researching, or archiving,
or undertaking with our smartphones.
And how you, it's what originally
got me into the topic of archiving,
social medial, using institutional
suppository technology as a basis
of doing that for researchers.
But, there are given the archivists
ethical frameworks are not
necessarily the same as
researchers' ethical frameworks,
which are not necessarily the same
as journalists' ethical frameworks,
what do you think something like the
Library of Congress should be doing?
I just turn it always onto
the Library of congress.
>> KATRIN WELLER: One rather,
one thought that just came
up to me was I remember, I
can't give the credentials
to who is doing that, but
there's a project for example
on archiving blogs by
Chinese, you know people
who write blog posts in China.
And they are being archived before
they get censored and thus disappear
because the Chinese
government takes them off.
So, there's a deliberate project
on archiving this kind of content,
which is being at risk
of being censored
and disappearing forever kind
of adding another ethical
perspective to that.
Or feeling obliged to help these
people to keep that stuff public
in some form before
it gets taken down.
And I think that's part of the
more archiving perspective on that.
When it comes to social media data,
I think you also have to distinguish
between let's say everyday user
accounts and more official accounts,
like the president, the
press accounts and so on,
where you have different sets of
ethical considerations to take.
And one thing I think that
maybe the Library of Congress
or other institutions involved in
this could start with this kind
of archiving the contents that comes
from official accounts by leaders,
political leaders, other types of
elites where you can clearly assume
that these people know
what they are doing.
They are doing it for
professional reasons,
and that kind of is the
first piece of the puzzle
and understanding the
entire framework.
But it's again, it's not trivial.
>> PHILIP NAPOLI: I think,
things start to get muddy here,
if we're truly focusing on
news media and journalism,
or more broadly because you know
I'm thinking about something
like the right to be
forgotten, right and if you think
about the layering of ethical
considerations that goes with that.
There is a news outlet
that presumably went
through the right ethical protocol
involved in reporting a story.
Now, if that story lives on in
Google's archive link essentially
and it lives on in this particular
news outlet's content archive
for longer than the subject likes,
the subject can ask to have it
of course petition for Google
to remove the link to the story.
So, there's an interesting, you
know presumably ethically there's a
reason why the archive of this
story should not be as accessible
as it currently is
essentially, but yet the right
to be forgotten has
never yet been extended
to the original reporting
outlet if I understand it right
that you can't go to the news outlet
and say you have to take this story
down because I am you know,
I haven't, you know we went
through various case
studies in one of my classes.
Well, I haven't you know
beaten my wife in 15 years,
so how can you still have
this story accessible?
I mean that's a real example.
But the fact that nobody is claiming
that that person has a right
to demand that the news
outlet remove that story
from their archive, but there is
some right that they have for Google
to not make it easy to find it.
You know I'm not answering
the question here,
except sort of again
showing you how these layers
of etherical considerations that
sort of cascade over each other.
But yet when you get to these
questions around the issue of yeah,
you know this was my Facebook page
and all my postings and my comments
and what right do people have to
store that and make use of it,
that's I mean to me the level of
overlap is almost minimal there.
They are very different
kettles of fish.
>> MATTHEW WEBER: We may need
Jefferson to come back up and talk
about legal requests that the
Internet Archive has received.
Although he's shaking his head no.
I think we had another
hand directly behind Les.
>> NICHOLAS HILAGIS: Hi.
My name is Nicholas
Hilagis [assumed spelling]
and I'm a research assistant
in Southampton working
on [inaudible] and a lawyer.
So, we talked a bit about
the legal aspect of things
but to me there's also
a research ethics aspect
which is completely different.
Archiving even for historical
reasons is going to be mainly used
for research purposes and in
any university in any study,
in order to get ethical approval,
you need to a, inform your subjects,
and b, allow them to have the right
to withdraw their information
at any time.
So, when notification happens, as I
understand currently in archiving,
but since we also archive
comments of people that have gone
to a webpage and they are not even
aware, it's up to the discretion
of the publisher to notify them,
maybe we should start thinking
about changing archiving practice to
allow compliance on letting subjects
to withdraw their consent, I would
just like your opinion on that.
>> MATTHEW WEBER: I'll make a
quick comment on that which is
that this is one of the thorny
issues as an academic researcher
that we muddle through with
institutional review boards
at US institutions, and IRBs
which is what we refer to them as,
do not know the answer
to this question.
And in fact in most cases,
don't even want to wade
into this water right now
because it is such a thorny issue.
The attitude more or less to date
has been well, the comment is put
up there, you knew you were
putting it on a public site so,
we're just going wave our hands,
but I think long term that's going
to become a more complicated issue.
So, my short answer is there is
no answer, at least in the US
at research institutions right now.
But it's important.
>> RAMESH JAIN: This is not an
exact answer to your question
but in 1994 David Brin wrote a
book called "Transparent Society."
And in that book he argued
that in future there are going
to be censors everywhere which are
going to recall each and everything.
And he argued that there is
certainly no point discussing
whether censor should
be put or not put.
Okay. Because that's going to
happen independent of everything.
What needs to be discussed is
who has what information and how
that information could be used.
And then he argued
something very interesting.
He argued that from the beginning
of technology we always argued
about the convenience
versus privacy.
And in the long-term
convenience always won over.
It's starting from use of checkbooks
to credit cards, to
internet shopping.
These are the examples that he used.
So, the point with
respect you are capturing
where these things are
being put etcetera is going
to be very difficult to stop those.
Particularly with the kind of camera
and all these things
that are coming in.
The correct [inaudible]
society will decide is what kind
of information can
be used ethically,
legally where it could be
used and things like that.
And this is really going
to become more important
and slightly different
context with video devices.
Because everybody's
collecting this information.
That information is your
very personal information
and that's been going to the
cloud and is being collected.
So, how are you going to use
that information Google
has access to that.
So, people aren't thinking
always putting
in the architectures
for those things.
So.
>> KATRIN WELLER: Maybe
two additions.
First of all, I live in a country
where there actually is
no equivalent of IRBs.
So, German universities don't
run through that process.
Yeah.
>> RAMESH JAIN: I should
work with you.
>> KATRIN WELLER: We've got other
legal frameworks to cover that,
but I mean that's not the standard
that as a researcher you have
to file something through
an IRB process.
Well they've got some forms
they have to fill out.
But it's not an institutionalized
board such as IRB, which is there.
You have to check some boxes
on a form and that's the case
and there are legal restrictions for
what you can do and what you can't.
But it's working quite different
than in other parts of the world.
The next thing is that I
don't believe that kind
of just getting clearance
through IRB solves everything.
Because on the one hand side, that's
again more of a legal issue kind
of if you just need this in
order to get the allowance
to continue some processes.
And if you need someone to sign
kind of an informed consent form
without really being informed
or understanding what he
or she is signing up
for in that moment,
that doesn't really solve the
ethical issue around that.
So, it's, yeah really I join in with
your answer, there's no easy answer.
>> PHILIP NAPOLI: I think what
makes it interesting in this context
at least for the kind of work
we do on new and journalism is
when it comes to IRB, it's
human subjects, all right
and this, okay how do we define?
When are you a human subject?
To figure is it your digital
expressions online does
that you know at least in part of
the research we're doing right now,
we're not doing research that
actually reaches the level
of studying human subjects,
even though we are producing the
outputs of human effort, right?
And so, you know what's
the threshold for counting
as a human subject is
kind of interesting
when you're studying digital media.
>> SPEAKER 4: Do you think that the
role of libraries might expand to,
or even if it currently does,
you know providing people
with more digital literacy and kind
of the tools that they need to make,
to actually do informed consent and
to understand what their content
that they're generating online
is actually how it's living
and what it's actually going
to be potentially used for.
Is that something that you think
libraries should be undertaking,
like that education, or is it
you know the responsibility
of the, you know parents?
Or society in general?
I don't know if you want
to speak to that, but?
>> KATRIN WELLER: Well,
I think information literacy
is a very important concept.
But it also adds another level
of making everything more
complicated in that sense.
Because as people get more
literate in using different tools,
they will put on new practices
of how to use these things.
So, it may probably turn
to different channels
for talking to one another.
So, for example once you figure
out that Facebook is doing certain
things, you will use what's it
for a specific task and then
you move on to some other tool.
And, so on and so forth.
So, it's really a very
dynamic field.
It's a moving target
that you try to capture
and it's constantly involving,
I'm not sure whether
libraries should be the ones
to be in that position.
They should be aware of what's going
on, they should be able to take
up that role when they're
asked for it.
But they shouldn't be the only ones
who should take care of that role,
who will be responsible for all
this, but I think the entire dynamic
on the web is like users taking up
practices and establishing forms
of how to communicate
with each other.
Platforms changing their
futures according to the users,
putting on these new practices and
everything goes on, and on, and on.
And that's really something
we need to be aware of
and we keep considering
in these kind of efforts.
>> PHILIP NAPOLI: I mean my
short answer is just going
to be all of the above.
I was waiting for the issue
of digital literacy to come
up in this conversation
because I would love to know
from a research standpoint
how variations and levels
of digital literacy
amongst individuals,
you know to what extent is the
relationship between that even
at the organizational level and
what we might call chilling effects,
right is you know and that how much
does that input what gets produced?
Ultimately then what gets archived.
I mean from a research standpoint,
just we don't know enough there yet.
>> MATTHEW WEBER: I think
we have time for one
or two additional questions.
One over here.
>> SPEAKER 5: I know that the title
of this panel is Saving Media,
but my question is about taking
it into the future really
that is not just saving it
historically, but your example
of an article about somebody that
he beat his wife 15 years ago,
wouldn't it be nice if, for example
he was, I mean it was proven not
to be correct, that that particular
article would be annotated
by the newer information
about the case?
Similarly, we read all kind of
things, vitamin D is good for you
and then five years later,
don't take it you will die.
Okay? Now I think that we have
an opportunity in journalism
and either social media to
be try to do these linking
to more recent information
on that particular topic.
In other words, a [inaudible] is
very, very important and we can try
to use the dynamic web to represent
the current state of knowledge.
>> RAMESH JAIN: I like
the question a lot
because what it is
also emphasizing is
that in our discussions
today we are not doing,
we have talked very little
about the importance of time
and the importance of location.
And both those things are extremely
important because based on time
and location you can
decide many of these things.
And that we are going forward,
you allow all the tools
that will link all these
things very easily.
And that's happening increasingly.
>> MATTHEW WEBER: I think as well
as we are starting to become better
at working with the existing
archives that we have.
For instance, just
reflecting on the datathon
that we held over the past two days.
As we're learning about
ways that we can work
with the data I think we will start
to slowly answer these questions.
By analyzing the past
we'll know more
about how we can tie together
bits of information in the future
to create a more coherent
longitudinal archive.
Personally, as somebody who
researches on news media,
that's something I would
certainly like to work towards
as we continue tinker with the data.
>> SPEAKER 6: So, maybe a two-parter
if I don't get in trouble for that.
I guess, one I guess to go back
to the ethical considerations,
I think the concepts of public
and sort of capital being public
of you know I might be
publicly discussing something
with a specific community that
I'm involved in, maybe a community
of researchers, but as soon
as maybe someone studies this
or a journalist decides to
publish about this, you've now sort
of raised a profile of what
I was discussing in there,
I think we're, what are the?
How do we sort of update
our conceptions of harm?
Because IRBs and most of the
institutional review boards
across the world, they're sort
of legal frameworks to
protect universities.
They're not ethics,
necessarily ethics bodies.
Because you know of
how they were founded.
So, you know one is how do
we come together to sort
of update our conceptions
of what does it mean
to collect someone's geographic
information that's attached
to their photos that they're posting
on Flickr and do this in mass.
Like with our occupy datasets, we
can figure out, in my lab we figured
out where followers live, you
know where they were protesting,
we say the use of turning it
on and off as a protest tactic.
We saw the use of deletion
of the protest tactic.
So, on one hand we want
to accurately reflect what was
happening on the other time,
but that's also what about
the harm that we could cause,
you know the Oakland
Police Department would
like to subpoena our
data, they did not,
but I mean if they could
subpoena our data set.
They could, you know find
some of the protesters
that they might like to arrest.
You know we have how do we
understand those conceptions
of harm, I guess I'll
stop at that one,
because that might be big enough.
RAMESH JAIN: To me, and the view
that we are discussing here
are really very complex.
We want to understand them
in two different levels.
Number one is the technical level
and then from the societal level.
Which is more politics
and ethics related.
So, in the archiving component.
We have to see carefully what
could be done for example lots
of views today about what
information should be shared,
what should be discarded.
So, is it based on the politics
or ethics, or is it based
on the information content of it?
We all know that we
do image compression,
we do video compression.
Are we going to do the compression
of all the data that's coming
in based on only that kind
of issue, or we are going
to start making the decision
based on political ethics.
These two things should
be kept separate.
Because the technical
development is going
to take place based
on what could be done.
And then we modulate that
by societal standards.
Because these societal standards.
Because these societal standards
are going to be governed
by the current socioeconomic
and political situation.
Today, what is allowed, tomorrow it
may not be allowed and vice versa.
>> SPEAKER 6: The technical
objects, the technical system,
then that's forever there and you
can't go oops I want it back, right?
I mean so?
>> RAMESH JAIN: Most of
the technical systems
when they are developed,
then the society decides.
For example, Facebook, Twitter,
and all these things
are banned now in China.
Because the society decided there
that it's not in their interest
to allow users to use those things.
But the technology they
have developed based
on the technical thing,
their own systems
which are sometimes correctly
used sometimes are being abused
by people in government.
So, technical issues
and the political
and social issues are different.
At least that's how I see it.
>> MATTHEW WEBER: Wow.
>> SPEAKER 6: We disagree,
but respect that.
>> MATTHEW WEBER: On that
contentious but interesting point,
I think we may draw a
line and end this panel.
I want to thank everyone for
joining us up here today.
[ Applause ]
Wendy for closing remarks.
>> DAME WENDY HALL: So, for
closing remarks, this is where I get
to say what I want to say.
All that we've been talking
about today has been driven
by a revolution that started
when people started talking
about machines, and computers.
But in terms of the real impact
on the world, of the internet,
it started with Tim Berners-Lee put
the first website up Christmas 1990.
And in that very short space of time
our world has changed dramatically.
Lee talked about this on
the panel this morning,
it's hard to think back in history
what you can retrospectively see
when things hugely changed,
industrial revolution and things
like that, and agricultural
revolutions
and the different you know ways
of measuring what we
changed in society.
But this has been something
that's happened in our lifetimes,
which is probably the first time.
And, we, as I said this
morning this system
that underlies is still
very fragile.
And we the thing we've been talking
about today it's so very, very,
very complex that it's
almost impossible to grasp.
What we've done today is
just the tip of the iceberg.
We could have had days talking
about this and of course
in the future people will.
The you know we have when you
think about the, oh, sorry I've got
to do this so I can see it
here, but when you think
about the way libraries
have evolved.
And somebody earlier,
I can't remember,
it was lovely, was it Vint?
Talked about the Library
of Alexandra
and you know way back in history.
And the big libraries we have today
have been evolving for centuries.
And the material in them
is centuries old as well
as decades old, and
as well as today.
And the sort of ways that society
has evolved to you know people come
into reading rooms and there's
these cataloging indexes,
which the wonderful Dave who
I saw sneak in somewhere.
If you want a good tour,
tour the library talk
to Dave Brunthin [assumed
spelling] over there.
Took us around the catalogs,
there's some wonderful
gems in the catalogs alone.
These themselves are
amazing artifacts now.
Right? The catalogs.
And I've been privileged to be
in this wonderful Kluge Center
where people come and do research.
And most people in the Kluge
Center read books and write books.
I don't do that.
I network, but there you go.
The thing is that this has
taken centuries to evolve
into these types of systems.
And I'm very well aware from my time
on the board of the British Library,
I was the non-exec directory of
the British Library for eight years
and then being immersed in this
library for three months how much
of the library is to
do with this world.
Right? Because it's only 15
years since the library had
to start thinking about
the digital world, really.
That sort of period.
Most of the staff who work
here, most of the operation
of the library is to
deal with the old world.
And that's going to
continue for a long time.
So, there are major issues for
people who manage these organization
about transitioning
from one to another.
And the other thing that's
happening is that the,
as I travel around the world, and
I see and I talk, I mean just,
well I don't have to travel around
the world, I come to Washington
and I have talked to many
different archivists.
They've all been in this room.
There's the Senate archivist.
There are people, the National
Archives, the Congress,
the all the university
archives, the library here.
Just in this one town, there
are any number of archives
and there are people
grappling with this issue
of moving into the digital.
You then scale that up to all
the libraries in the world
and then you scale out,
you add in my actual world,
which is the university,
research labs.
And we are as much as you are
harvesting stuff off the web all
the time.
There's huge amounts of
duplication of effort.
We're still losing stuff.
And what we're actually ending
up with is a very complicated
jigsaw puzzle
for future historians to sort out.
Now, Vint talked about this in a
much more eloquent way than I can
about the complexity of this
amazing thing that we have created.
I mean Tim Berners-Lee
gave us the protocols,
but the web was created by us.
We create this content.
I was struck by a comment from
Abbie when she said, oh she was sort
of cursing YouTube you know
for it coming into existence.
Why did YouTube come into existence?
Because all of the
world wants to look
at cats doing stupid
things on videos, right?
Who would have predicted?
But you know?
YouTube exists because of us.
No technologist said, you will
put your video on YouTube.
They provided the infrastructure
and we put our videos on.
So, now of course it's
become a big business
and it's completely
changing the world
of entertainment, advertising,
everything.
And I mean my mentor sitting
in the front row there,
Professor Ramesh Jain, who you
know I met at Michigan in '89
and he was the founding
father of multimedia
and you know we have been watching
this happen for a long time.
And as he said in his talk,
it's only today we're beginning
to realize it isn't just
text we've got to sort,
it isn't just you know data,
it's all this, well it is data,
it was the end of it
it's all bits and bytes.
But it's pictures, it's
videos, it's audio,
it's a huge amount of information.
And we have invented the technology
to enable us to all of us who are
on the internet and
have a mobile phone,
we can all create this
stuff isn't that wonderful?
You know and that, I mean
what do you think about it?
During the analogy of the old
days, we wouldn't expect a library
to have kept every little letter
that was written in America, right?
You'd expect a library,
a national library
to store the important letters
that were written in America.
Not every single personal letter.
And not every single
personal postcard.
You don't, some people
collect postcard,
that's a different madness.
And where have postcards gone?
To Twitter.
I mean you know.
Largely you don't send a
postcard anymore, you tweet,
or you Instagram, or you
Snap Chat, or you whatever,
you put it on your Facebook page.
We don't have to collect it
all and yet there's a mess.
The mess isn't just what we collect,
it's how we collect it and the fact
that we currently, everything's
in silos and we've got bits
of the jigsaw puzzle in silos
and they're all different
full of standards.
Right? There's no common way of
these jigsaw pieces being cut.
So, that they mesh together.
So, I look back to that
table Vint put up on slide
with the hieroglyphics on it
and I think you know how long
it took people to understand
that Egyptian language and
put the tablets together.
And as someone else here said,
I can't remember who it was now,
I think it was the data panel,
someone said the most
important thing is the context.
Given the context,
I think it was Vint
that said it put the
context of the data up.
It's all about the
semantics of the data.
Because if you don't do that, people
in the future will not be able
to understand what it was
about because they won't
understand the world we lived in.
They won't understand who
Snowden was or, I have no idea
who that Judge was that
you all doing the thing
about the court okay,
but the supreme group.
I don't know who that judge was.
I can look him up on Wikipedia, but
how long is Wikipedia going to last?
Who's preserving Wikipedia?
You know? I mean I'm sure
they are, but can we assume
that the people running
Wikipedia will be in existence
in 100 years' time or that a
system running Wikipedia will be
in existence in 100 years' time.
Certainly, the people
doing it today won't be.
Now, these are assumptions
we make that just are not,
that are totally foolish
assumptions to make in this world.
So, the Library of Congress,
we've you know we've heard
about the millions of
resources we have now that
and a huge amount of time.
Just imagine if a library
ingested those numbers of books
in 15 years, real books you know.
This is the scale that
they're working at.
That's a lot of websites.
Just this library, and that's
replicated all over the world.
And you know you deliver
an API service.
How do people find out about that?
And how many people know what an
API is and what they can do with it?
Right this is the trouble.
It's the same that's
happened with open data.
All sorts of people put data bits
out there and nobody uses them
because they don't
know how to use them.
And Jim said to me earlier when we
were sitting listening to someone,
what we've got to do is write
the tools to enable people,
just regular people, but
particularly researchers
who aren't computer programmers
to be able to do things
with this data and
that means chairing.
It means doing this as a community
because you know otherwise we're
just going to waste a lot of time
and effort or duplicating
and getting nowhere.
So, we've, I'm going to
tell you a bit in this talk
about a project we've done.
This is the website I trust
which you've heard a bit
about sponsors this project.
And we've done the building at
Southampton, of a simple tool
to enable people to find archives.
Find out what's been archived.
Because at the moment we
have no clue who's doing what
around the world.
We hear about it at symposiums
like this, we hear about it
in conferences and also
when students do projects.
Interns come in, PhD students
come in, they do projects.
They do wonderful things with data.
They leave and it just goes.
You know their laptop goes,
the server goes, the URL goes.
And so, we've been building this
thing called the Web Observatory,
which I'll explain what
that means in a minute.
It's all about trying to make
data about archives and data
about research projects that are
being used in research projects,
the knowledge about that
available to people.
It's effectively about
creating data catalogs.
And being able to discover them.
So, that my Ramain [assumed
spelling] who's sitting
in the audience somewhere
produced this slide so that instead
of putting out an API, you
put out some simple metadata,
which we talked about a lot
on the panel this afternoon.
We are currently using
shema.org which is a sort
of watered down version of RDF.
But, at least it's a standard that
we can propose, as you know the idea
that everybody who's doing work
in this type of work agrees
to a common description
of the datasets.
Not, the data, we're not talking
about formats of the data,
just the descriptions
of the data sets.
Who owns them, what they're
about, what you can do
with them, who can access them.
That sort of level.
Oh, I'm sorry I knew
I'd confuse you.
I'm looking here because
I want to be able to read,
you can't read from there.
All right so to go back, this is the
web observatory that we're building.
We're using schema.org.
It's all about metadata, about
datasets, so that everybody can flag
that they've got some datasets.
And so, you can put out an API,
absolutely because you want people
to use that data and
build tools to use it.
You also put out the
metadata about that dataset.
And if we all do that in common
standard then wonderful things
can happen.
For example, and this is
something that we're just beginning
to experiment with you can begin to
search across different datasets.
And here's this is actually it's
a beater, beater, beater thing,
or maybe alpha minus thing.
It is a we're just developing it.
So, it's like a Google for
data, or metadata anyway.
And when Romain did this yesterday,
he typed in pollution, I think Kate,
we were there in Dunk 'n
Donuts at the time weren't we?
Oh, there you are right?
And he typed in pollution and he
gets back from, this is searching
across all the observatories
that we currently have registered
in the system for who's got
datasets about pollution.
It could be a Bolder, it could
Zeke, or it could be Trump,
it could be whatever you want.
And so, he searched, but you can't
read it but the first one that comes
up is a dataset from Southampton
and the second two are datasets
from the Library of Congress.
Now, you're not actually issuing
this metadata at the moment.
I hope you will, but
we fudged it for this.
No, we didn't?
Oh, it's real.
It's real.
It's in the Southampton Observatory
and these are datasets that we use
for the datathon, but it's
only the metadata, all right,
we haven't sucked the data out, just
the metadata that says the Library
of Congress has these datasets.
So, this is, imagine you can go to
something like, well it won't be
like a Google, because
it won't be as big,
maybe Google will run the service
at some point in the future.
The idea is you just,
you're searching
across the world's
datasets of web archives.
Just imagine being able to do that.
So, here's another one where,
oh yeah you highlighted,
is that the same one,
the pollution one,
you highlighted Library
of Congress on this one.
Okay. Now, imagine a
world, where, so here,
we've typed in the world elections
and in fact it's UK elections.
And we come up with the dataset
that we are hosting at Southampton
about last year's UK election.
We harvested from Twitter.
And the datathon, the group
that worked on this data,
and they have the US
election data from the Library
of Congress they produced a
dataset with the work they've done
and that dataset is now, well it's
pointed to by the observatory.
So, here's the metadata about on
there that says, you can't read it
but it says this is all screenshots
of a live search that says,
this is the data that was
produced after this datathon.
And what I want to do is every time
students or anyone does a project
with some of the data from anywhere,
like from the Library of Congress,
or the British Library,
or Southampton University,
or George Washington, or Rutgers
University, or RPI University.
If you're doing something with data
that you got from somewhere else.
When you produce that data,
you then put the metadata back
to say what you've done
and where the results are.
Right? So, you build up a community
of people working in this area
and we know who's archiving what.
So this, the idea of the
observatory came from the physicists
who are very good at sharing data in
order to interpret what's happening
out there in the heavens.
The thing about that is
that's the physical world
and you don't change it by
observing it, we don't think.
With the data we're doing,
there are all sorts of issues
of using data that's about people,
as we have touched on a bit today.
But the idea is the same.
It's that as a global community
we've got to share this data
to enable people to
replicate our experiments
to reproduce experiments, to
put metadata on top of metadata
to add some context to the
datasets to help people
in the future interpret
what's going on.
So this is about sharing.
There's a diagram, I think
Romain came up with this as well,
many years ago now, if that was
the web browser, that was the web
in 2003 I think, which is often
used to illustrate the connectivity
of the web, and it's
all about observing it
and experimenting with
the data on it.
So, just the Web Observatory, if
those of you who are techy want
to look at websites.org, which is
hosted by the trust and you can look
at the index of the Web
Observatories that exist
at the moment and how
you get into being one.
It's very simple.
And there's a list of, that's
a screenshot of the list
of web observatories
that exist at the moment.
Jim, we need to make yours live
again, I think RPI is the top
and it needs, the datasets
need to be live again.
The trouble is if people
don't use this,
and you don't have someone curating
these datasets, they do stagnate.
This does need, this
is not something
that just happens automatically.
But if you go to for example the
snap group at Stanford, they flag,
they have, this is an
amazing group up there,
they flag that they have datasets
they've been collecting for years.
And flag them here so that
others can see what they've got.
And we can put them
into the search engine.
So, hopefully in the
future we're going
to be creating a distributive
catalog
across web observatory sites.
That's the plan, the
common standards.
The idea is you host your data.
It's your data and it's
all about providing,
telling people what
they can do with it.
You can tell people it's there,
but not allow them to access it,
you know on the web,
without going through you
without signing whatever
forms, whatever terms
and conditions they have to take.
But it's all about
cataloging what you've got.
But as a community and not just
as individual websites produced
by the individual institutions,
that's the difference.
So, then you have communities based
around different web observatories
and then linking between.
This is the Southampton one, and
if you, this is live on the web.
And if you go into the datasets,
you get lists of datasets,
and you get actually you get
told, we've got access controls.
Whether you've got access
or you need permission.
And if you look at the
apps you get a list of the.
Now this is the thing that
these take a long time to do.
This is coming to Katy's
material right?
All of these apps are
visualizations.
That's much harder
to share than it is
to share the metadata
about the datasets.
And the amount of effort that
goes into building those apps.
We don't want to be replicating
that all around the world.
We really, really, really need to
share those tools wherever possible,
if they're open source, or open
access whatever, we really need
to be able to share them.
And this is about telling people
what you've done with your datasets.
So, for example, you can't
read it but you can look
at this, it's all online.
There's datasets here about
Wikipedia live revisions
of Wikipedia we have that
constantly running in Southampton,
we have a screen up with
who's editing Wikipedia
at that moment in time.
So you've got streaming data
and then that's all stored
in the observatory and flagged
and then you get these sort
of you can actually give people
that, I won't that is a video,
you can run it as a video and see
who's doing what with Wikipedia.
And you can get that data, you
can look at it streaming live
or you can get the
past data as well.
And of course, the wonderful
thing about Wikipedia
for researchers is it is all open.
So, with the idea is we have web
observatories all over the world.
And they're all linked up
and we have the Southampton,
a couple in the Southampton, RPI, we
just ones being set up in Bangalore,
with a lot of industry data actually
about stuff, telecoms I think.
And then there's the new
one in South Australia.
And the idea is that
libraries have them too
but actually this is the modern
version of your catalog I think.
And that's the plea today, is
that we develop common ways
of sharing our catalogs
about our datasets.
In order to as I said
earlier describe the content
of the observatories, or what's
in that datasets, we're
using schema.org.
This was developed by
Jim and his group at RPI,
it's in the w3.org list of schemas.
Anyone can use it.
And that really is a very
simple just a description
of what's the name of the project,
what type of data, who owns it,
what access you can have.
It's just at that level,
but that is enough
to enable us to do
initial searching.
And we can put more and
more descriptors in and then
over time we'll be able to go
semantic, which is what I want to do
with Stanford, is to link these two
things together, because it's only
when you really get rich
semantic descriptors
that wonderful things
begin to happen.
So, we have a small ambition
here to map the digital universe.
That may be MP incomplete.
Anyway, but who knows.
The idea is to get a better picture
of the digital universe
than we have today.
So, less broken jigsaw puzzles and
more bits of the picture emerging.
And I just want to almost finish.
This is my last but one slide,
with a video that we produced
at Southampton, which
may explain some
of this a bit better than I did.
And the voice, if you
play it in a minute,
the voice in this video
is Julian Rhin Tutt.
We were so lucky to get him
because they didn't charge us
very much for making the video.
Julian Rhin Tutt if you
want British television,
he's always in as a character actor
and he happens to be the brother
of someone I went to school with.
But anyway it all links.
Links. I love links.
Could you play the video.
[ Music ]
>> JULIAN RHIN TUTT: [Background
Music] The Worldwide Web.
Pages, people, data.
Lots of data.
Generated and stored by
people, businesses, communities.
Imagine a world in
which your data combined
with other data could
change things dramatically.
Government, health, education,
the environment, and in business.
To unlock the potential
of your data,
we've developed a Web Observatory.
A system that locates and describes
existing datasets around the world.
It has common technical
standards, safeguards,
collection, and analysis tools.
It allows people to observe the web
and use the intelligence it holds
to unlock new opportunities.
You chose how your data
is used and by who.
Well it's free or has
to be paid for.
But until you put it out
there, you won't have any idea
of what it can do, how
it can change the world.
So, join us we can show you how.
The Web Observatory, unlocking
the potential of data.
[ Music ]
>> DAME WENDY HALL: Thank you.
[ Applause ]
I should give credit to Thenasis
Jurapanis [assumed spelling]
and Joanna Luis for doing that.
But this is not a commercial
company, right?
This is just talking
about this project
which is a community
project, or open source
and open development tools we hope.
And we can't do it on our own.
We need to do it together.
So, it's a plea for everybody
to put their metadata out there
in the schema.org format
and tell us they've done it.
And just to summarize today,
this is my last slide.
The center of what we've been
talking about is this web, internet,
whatever you want to call it.
The thing that is incredibly
complex, as Vint said.
Very, very interconnected.
As, Ted Nelson who's another
great mentor of mine mad as it,
oh no get quite mad, but a mentor
of mine said everything is deeply
intertwingled and boy he said
that in the seventies and
boy have we proved that.
You know given, he didn't
quite get Xanadu off the ground
and he says two cheers to
Tim and the Web, not three,
but you know the web has shown us a
way to do this and it's the center
of what we're going and I love
what Vint said the morning,
the self-archiving web that's
something we really need
to try to build.
It's very complicated.
I love the term digital
vellum, that resonates
so much in a building like this.
You know this is really, this
is what the libraries need,
they need digital vellum.
And that's what we need
to provide them with.
All right?
They can't do it, we the community
have got to do that for them.
We need to know who's archiving
what, otherwise we get huge amounts
of replication, duplication,
wasted effort.
That's where I hope something like
the web observatory will come in.
And for those of you
who can understand all
the three letter acronyms,
the doi stuff is really important
because in the web observatory
at the moment, we're focusing
on sharing metadata
and we're using URLs.
And as Vint said, URLs
are inherently unstable
because people don't,
people move stuff around.
And so, the work that Bob Kahns
and CNRI are doing on doing on doi,
as complicated as it feels, you have
to have that sort of system going
at some point in order
to do the preservation.
So, on that note, I'm
going to finish.
I'm going to say thank you.
I think Janice is actually
going to close.
All right.
I'm just going to say thank you
to all the people who helped here
at the Kluge Center, the people
to be organizing, Jason who's been
at the back you can now have a
rest Jason I know you've got an
event tomorrow.
The team who produced
the food, ran the mics,
the technical team,
thank you very much.
All the behind the scenes people.
And then the front up scene people,
all the panelists, the chairs,
the people who ran the datathon,
thank you so much for everything.
And thank you to the Kluge
Center for letting me be here.
Thank you.
[ Applause ]
>> JANICE HYDE: Wow,
we've heard a lot today.
And I want to thank
all of the participants
for really their thoroughly engaging
and thought provoking presentations.
I don 't know about you,
but I need to look at some
of this again and reflect.
And so, I just want
to remind all of you
that we did record the
presentations and the will be
up on the Kluge Center website and
the Library of Congress' website,
might take a couple months
but they will be there.
And by then we'll all be ready to
go back and look at all this again.
So, I would especially like
to express our gratitude
to Dame Wendy Hall for
organizing this event.
[ Applause ]
And really for serving
as the catalyst
for this great exchange of ideas.
And during her tenure here
at the Kluge Center she has really
generously shared her abundant
energies, talent and knowledge
throughout the institution
and across the hill.
And I want to thank her for truly
embodying the Kluge Center vision.
Thank you [applause].
>> This has been a presentation
of the Library of Congress.
Visit us as loc.gov.