I chatted recently with Olivier Thereaux, Yves Raimond (senior
technologist in R&D), and Silver Oliver (data architect) of the
BBC about the Web, publishing, and linked data.
Ian: The BBC is prolific and large. How do you
view yourselves?
Silver: The BBC is primarily a broadcasting
organization. Content is developed or commissioned within different
editorial domains (such as News or Music or Sports) then
distributed through diverse channels (TV, Radio, web, apps, etc).
This fragmentation exists also on the web, with development of
individual sites being largely delegated to dedicated teams.
Ian: How do you move beyond silos?
Yves: We have a lot of data that we are now
using to draw connections among various BBC TV and radio programs
and entities in other domains, like music or nature. We also expose
the corresponding data. For example the programmes site exposes
data
views giving details about all the music tracks played in a
given radio programme, and those details link to (and draw from)
artist profiles on the BBC's music site… which themselves are
also available as
data views.
Olivier: We also reuse data that's available on
the Web (e.g., from musicbrainz and wikipedia). Because the public
is curating the information they can update it more rapidly than we
could on our own. In a way, the Web is our Content Management
System.
Ian: What are you using to aggregate and expose
the data?
Yves: For the programmes and music site we use
a relational database internally but then we expose the information
in RDF.
Olivier: And we benefit from the ways that
people have innovated around the RDF data we expose. When people
play with the interfaces and massage the data, we can build on
their experience.
Ian: Why not use RDF internally?
Yves: I think the main reason is that the
people who originally built these sites site were unaware of RDF,
or were concerned about using an unfamiliar technology on such a
big project. But we use it with other projects.
Ian: How has your uses of data affected
reporting?
Silver: In the past our editorial efforts have
been captured in whole HTML documents. This causes problems for
reuse in new data views and across platforms and applications
(including IPTV). The key is in working with existing editorial
workflows to capture a sub-set of machine readable information. In
its simplest form this might be a byline and small number of tags
the story is about.
Ian: How do reporters use the data to make
connections between stories?
Silver: Connections have always happened, but
it didn't scale. Linking between sports and news was a manual
process and reliant on a journalist's knowledge of BBC output. But
now we have rich data models behind the scenes. These models help
the BBC editorial staff represent their understanding of the world
and our audience's interests, and let us make connections in a
scalable fashion.
Olivier: The data is a substrate that
pre-populates a lot of the site, and then journalists can focus on
the stories and not re-entering the data bits.
Silver: In sport, for example, we pay for the
sport data (fixtures, results and statistics) then we write stories
about match reports, and tagging ensures that everything gets
linked properly. That's
how we built the sites for the
2010 world cup or the 2012
olympics.
Ian: Do the reporters add data to the system
directly?
Silver: Yes, we ask them to tag the stories
they pull together so that we can put those stories into different
contexts (or aggregations). We were quite happy to realize the
natural curatorial process was already happening, we just needed to
give people a way to capture data.
Ian: You mentioned buying and using data from
various sources, including commercial ones. Do you make use of data
provenance information?
Yves: We need to be very transparent about
where our data comes from. Our reporters, partners, official
organisations, sometimes our audience too.
Olivier: There is an interesting tension
between making use of provenance information and ensuring user
privacy. These days people expect to receive personalized content.
To achieve that we make use of "attention data": what you watch or
like. We have been looking at how to guarantee that we uphold
privacy while at the same time asking for the minimal amount of
information to tailor the best experience. That's probably less
about "Do Not
Track" and closer to the spirit of W3C's older P3P technology.
On the other hand, we want to know whether information is reliable.
This is challenging for user-generated content in particular: who
is the user? how much do we trust them?
Ian: Do you think making provenance information
available to readers can help digital literacy?
Silver: We had an interesting debate internally
whether to include links from health stories to the journals that
published the original research. Some felt that readers would not
be interested in the links or would find the research complex.
Others encouraged the links so that the community could respond to
our articles with their own interpretations, including challenging
the articles from various angles. This, in turn, would generate
more discussion and perspectives from a much larger audience.
Ian: How did it turn out?
Silver: In stories about politics we have begun
to include links to relevant legislation. And we are exploring how
to extend the linking to pull in data from these sources to weave
into BBC story-telling. For example data about committees that
commented on bills, which members of parliament commented, and so
on. These data models allow us to make more connections among
stories, as we discussed earlier.
Ian: This sounds like a linked data
project!
Silver: Internally we have wholesale signed up
for and understood the value of linked data as way to manage our
organizational complexity. We will draw data from various sources
and use RDF to stitch them together. We can make use of the
information in ways we could not do before because it was either
too costly or unmanageable. Semantic Web technology is now core to
our strategy as an enterprise.
Ian: Have you measured cost savings by using
Semantic Web technology?
Silver: It's still too early to say. There were
costs associated with our initial projects, since we needed to
acquire expertise. But we have since been able to roll out highly
trafficked BBC content using Semantic Web technology.
Ian: Thank you all so much for your time!
since 2005.
