This is the presentation for BBC News Storyline data model part of the BBC Connected Studio briefing event for the BBC News Archive Connected Studio.
The presentation is aimed at providing context for the participants, so that they have a view of how content will be tagged, and how it will surface in BBC News.
News Archive - BBC News Labs presentation on Storylines, Topics & Tags
1. Storylines,
Topics & Tags
An explanation
of the BBC News Data Model
for the News Archive
Powered by BBC Connected Studio
(January 2014 : Belfast & London)
2. Jeremy Tarling: Senior Data Architect, BBC
News.
@JeremyTarling
Matt Shearer: Innovation Manager, BBC News
Labs.
@Completedespair
@BBC_News_Labs
3. linked data tagging
the audience can find and follow content
relevant to them.
topic tag: a person, organisation, place or
theme.
storyline tag: a storyline, or its component
events.
4. why do it this way?
it powers serendipity.
we organise by the “things” people want.
and connect with meaningful links.
5. context
Storyline is an open model.
BBC are now tagging new content.
the archive is not tagged yet.
6. topics - people
Bashar al-Assad
Thamsanqa Jantjie
Lara Clarke
Nick Robinson
a Person can have properties like
‘birth-place’, ‘birth-date’, and
roles like ‘President of Syria’ or
‘interpreter’
7. topics - organisations
an Organisation can have properties
like ‘address’, ‘website’, and can
be notably associated with a person,
place or theme
8. topics - places
Places can have a
latitudes/longitudes and parent
features (an administrative district
or country for example)
10. storylines
storylines are a way to link up and
present content to the audience as a
narrative
storylines are a special sort of
linked data tag for annotating and
aggregating content
storylines can be tagged with topics
18. before we finish
we can’t tag everything manually.
we can autotag topics,
and fingerprint into Storylines.
19. over to you
organise news content by storylines
link storylines together with topics
relate archive storylines to current news
surface unexpected connected stories:
“Death of Mandela” “Mandela’s life
sentence” “Thatcher’s foreign policy”
“Falklands War” …
Jeremy is co-author of Storyline, and drives the Linked Data strategy for BBC News
Matt works with BBC News Labs, and is running the 6 months to 2 years innovation horizon for BBC News.
tagging helps the audience find and follow content relevant to them.
It also helps some 8,000 of our Journalists find and share material quickly and efficiently.
topic tag: a unique ID (a URI) for a person, organisation, place or theme
storyline or event tag: a unique ID (URI) for a storyline or its component events
Events vs Storyline : can be a confusing conversation – think of events as “something that happens at a time and a place, and may have associated topics”
The power is then the connections that can amplify what we curate – massive serendipity.
We establish the “topics” people love, and focus on the Storylines of value.
Emotional connection, passionate connection.
This will help surface millions of hours of archive content, organised into the “things” the audience LOVE and WANT.
It’s open – not just BBC – Guardian, FT will be using it. SKY News are interested.
Also, at BBC it is not just NEWS – other areas are using the same topic model, so News Archive will surface there too.
theme preferred label is BBC ed pol – do we say “drugs” or “narcotics”?
In a se
similar to a hyperpuff in CPS
using the linked data s-p-o pattern we can use predicates to describe the nature of the relationship between content and tag-concept
these predicate/relationships can provide signposts to our audience in large content collections
storylines can contain other storylines
like chapters in a book
storylines can contain events
e.g. linking together live events into a wider narrative
There will be a degree of automatic tagging using concept extraction.
Speech-> Text -> CE-> CV & disambiguation -> Topic Tags.
Image-> face recog -> People -> CV & disambiguation -> Topic Tags.
This gives us Topics.
We can then associate to Storylines via Fingerprinting.
i.e. match topic [et al metadata] fingerprint to Storyline topic fingerprint.
So - Journlists to tell the Story : Storyline!! – and we hope machines can do the rest…
We have a good way to arrange the data.
BUT – the user experience and product feature possibilities are infinite!
We need your help!