The News Storyline Ontology wants to make it easier for journalists to deal with the world as they understand it – that is, in terms of stories and curated narrative arcs over world events. The ontology aims to be a generic model for describing and organizing the stories news organizations tell, while supporting whatever their approach is to handling those stories. It provides, in other words, a model for the news itself: how different stories relate to each other, how breaking news evolves and how the commonplace entities of people, places, organizations and events relate to news stories.
“The first benefit is for the news organization itself to organize things, but it also lets them put together web pages more flexibly and closer to the way we access information as humans,” says Jarred McGinnis, one of the authors of the ontology. Formerly head of research, semantic technologies at Press Association, he is now an independent consultant in semantics at his firm Logomachy Ltd. Fellow authors are Jeremy Tarling, BBC News data architect, and a former BBCer, Paul Wilton, previously technical lead, semantic publishing and now founder and technical architect at Ontoba, which specializes in semantic publishing.
The core of the ontology is the Storyline class, encompassing Storyline components like Event, as well as Topics that could associate to Storylines with entities from external knowledge domains, and Attributions to indicate the owner of a piece’s interpretation – accounting for provenance and accommodating for the natural bias that occurs simply because a story is told by a human.
In the real world, the ontology works something like this: An overarching narrative that ultimately becomes the Arab Spring begins as a single piece, when a man in Tunisia sets himself on fire, McGinnis notes. But then the tendrils start to expand or to envelope that: The Tunisian uprising becomes a storyline in itself with which the man’s suicide is an associated event, and then that chain explodes to include Egypt’s, Libya’s and Yemen’s uprisings, and so on down the line.
With the ontology in place at a news organization, “because you are curating sub-storylines it is relatively easy to say this story, in all these thousands of pieces of creative work, is part of something called the Arab Spring,” McGinnis notes. “It’s easier than to have to manually dig through and manually enter it in as part of the Arab Spring. It fosters this institutional memory.” And, it fosters a way to automatically evaluate the vast seas of stories for attributions, which systems can be built atop to help understand the personal or institutional angle or agenda slant of published works.
“It’s starting to create a semantic level at the story level,” McGinnis says. “It’s introducing semantics at a level beyond the basic entities of people, places and organizations.” Not represented in the ontology per se are tagging models to associate storylines to pieces of creative work, since every organization does that in its own way.
The ontology also includes Storyline Slots, to give flexibility in how narratives are ordered – by time or causality, for instance. Indeed, McGinnis says, he expects the ontology to be built on, extending through sub-properties more specific ways of relating sub-storylines to major story lines to fit a particular organization’s requirements.
The ontology also gives news organizations an insight into how their journalists’ domain expertises intersect. “Provenance comes in in that you get a data-driven view of what journalists are writing,” says McGinnis. That could be useful when the usual mid-east correspondent is on vacation but there’s another piece to add to the Arab Spring picture, for example; the ontology can help the organization find who else it has on staff who has done a lot of work on things that fit into the storyline. “That is a very exciting consequence,” he says.
The Ontology At The BBC
At BBC News, where some 400 to 500 news articles are published daily, the drive is on to prototype and workshop what a version of its web site that leverages the News Storyline Ontology will be like.
The big question for some time has been, Tarling says, “how might we use a Linked Data approach to organize our news content into the kind of storylines that we conceptually talk about, and provide users with more meaningful experiences by saying this content is about this person, place or organization and expose those semantic annotations as links, for instance.” Or, in a world of mobiles and connected TVs, and exposed user locations, serve up local news on those devices from BBC’s news app to personalize the experience.
With the News Storyline Ontology joined in the future to its ontologies for core news entities and events, for annotating the who, what and where, a more dynamic news presentation awaits. “Storyline kicks in when a story breaks, if it’s the first piece of news about it we mint a new identifier, or if it is part of an existing storyline as a new development, we associate our blog posts and news and art and video clips and tweets with that storyline identifier,” he says.
“And then if we want to take it to the next level of modeling, if it’s a big enough story now where there are sections or chapters, we have a nice way of automating what otherwise is a horrible, manual process and of presenting it to users in a more flexible way, by moving out of static HTML into more dynamic page presentations,” he continues. “And we can come up with interesting ways of representing the information [on the site], like timelines.”
A user interface tool for storyline curation would be a help, he thinks, one that’s built on the ontology and that writes data structures as triples, to improve exposure of entire storylines for use by journalists. “This will play to the ontology, to let the journalist say, ‘Now I have 15 events that are part of the Boston bombing, and I want to pick five key events to offer the user from the entirety of our output,’” he says. “Editorially, through the Storyline Ontology there’s a chance to curate that.”