The Storyline Ontology
Jeremy Tarling @jeremytarling
Data Architect BBC News
http://www.bbc.co.uk/news
semantic annotation
cwork:
CreativeWork
cwork:
tag
cwork:
BlogPost
cwork:
Gallery
core:
Thing
storyline:
event
storyline
ontology
core
news
ontology
rdfs:subClassOf
rdfs:subClassOf
rdfs:Domain rdfs:Range
cwork:
Video
cwork:
NewsArticle
rdfs:subClassOf
news:Place news:Event
news:Person
/Organisatio
news:Theme
rdfs:subClassOf
cwork:
etc...
cwork:
mentions
cwork:
about
rdfs:subPropertyOf
journalists „tagging‟
annotating (“tagging”) content
tool embedded into CMS
concept extraction/NLP for topic
suggestion
journalists accept/reject suggested
topics
pilot – location tagging
it worked…
except when big
stories broke
we write several
articles about the
same storyline
articles…
storytelling is
fragmented
manual linking
decays
massive amount of
repetition
from articles to storylines
develop a data model to describe a news
storyline and its topics
refine our content model to handle granular
updates (A/V clip, short-form, social media
update, long-form)
ask journalists to annotate („tag‟) these updates
with their storyline
collaborative model development
www.purl.org/ontology/storyline
www.purl.org/ontology/storyline
www.purl.org/ontology/storyline
www.purl.org/ontology/storyline
an example storyline
linking storylines
linking events
tag storylines with topics…
topics
topics are real-world entities, or things
people
organisations
places
themes
people
a Person can have properties like „birth-place‟,
„birth-date‟, and roles like „President of Syria‟ or
„interpreter‟
Thamsanqa Jantjie
Nick RobinsonLara Clarke
Bashar al-Assad
organisations
an Organisation can have properties like
„address‟, „website‟, and can be notably
associated with a person, place or theme
places
Places can have a latitudes/longitudes and
parent features (an administrative district or
country for example)
themes
Themes are the intangible things that we might
want to classify our content by: „smoking‟,
„unemployment‟, „health‟
health
unemployment
smoking
tagging with a topic
<:thing> :type <:video>
<:thing> :about <:David Cameron>
but is this video clip really about
the topic of David Cameron?
about-ness?
tagging with a storyline
<:thing> :type <:video>
<:thing> :about <:storyline>
<:storyline> :slug “Cameron EU statement”
<:storyline> :topic <:David Cameron>
<:storyline> :topic <:European Union>
<:storyline> :attribution <:Nick Robinson>
topics connect storylines
curation vs automation
two ways to present tagged content:
automatic aggregations where all content tagged
with that storyline, event or topic is included in a
chronological stream
manual curations where a journalist picks and
orders content in order to tell a particular story
automatic aggregation
anything with that storyline or topic tag
automatically surfaces it in that stream
this could be the default/out-of-hours state for a
storyline or topic page
less time-consuming, but no control over tone
and sequence
automatic aggregation
manual curation
more time consuming, but greater control
candidate content is manually selected for inclusion in
a storyline or topic page
attribution – manually curated storylines can be
attributed to a person or group (internally or publicly)
manual curation
demo?
production tagging with topics and storylines
live pilot of storyline tagging in the Midlands

Implementing the Storyline Ontology in BBC News