What’s the only and biggest startup in Istanbul, Turkey that’s working on the semantic web? The answer is Dakick, says co-founder and general manager Serkan Ünsal, a service for more than 100,000 members in Turkey so far that is focused on event ontologies and creating an events recommendation engine.
Users can follow people such as celebrities, venues, movies or other performing arts, and, based on the relations it’s building among entities in its growing database, be directed to shows that a followed actor is in that night on TV, for instance, or an upcoming concert featuring the work of a followed composer.
Users can select entities to follow on its service or Dakick can collect data from Facebook and Twitter accounts, too. An issue with getting people connected with the events that matter to the entities they’re interested in, however, starts with the fact that the providers sending their movie showtimes, tv schedules, exhibition, sports, trade show, conferences, performing art, and concert information Dakick’s way don’t always have a uniform perspective on data structure, if any at all, which makes interoperability a challenge.
“In Turkey content providers typically don’t know much about data structure,” says Ünsal, so Dakick is trying to educate them to move the needle here. “We are saying to all our content providers that if you agree on a data structure it will make search engines more meaningful to list some event data.”
Schema.org, Freebase and BBC ontologies are the starting points for Dakick to bring some order, and currently, it is doing a lot of the lifting when it comes to adding the data structure that helps things along when it comes, for example, to recommending users who follow the artist Van Gogh in the service about an exhibition featuring his paintings in a local city. While the outfits that provide movie showtimes and TV listings have gotten on board the structured data train, exhibit, trade show and concert providers are lagging. “Since you follow Van Gogh you’ll want to try to see that exhibition. For that reason we should relate his paintings with that exhibition. For that reason we should structure the exhibition data – exhibitions are composed of paintings for instance, and paintings are painted by artists, and you can follow that artist, and so on,” says Ünsal.
According to Ünsal, dakick currently counts up 335,065 celebrities; 66,961 movies; 2,759 venues; 16,654 music bands; 47,282 tv programs; 5,198 sports teams, and 4,075 performing arts. It assesses about 5 million triples for 1,000 movies alone. If you follow Brad Pitt, for instance, you’ll get an alert when Fight Club is on TV. With its ability to draw relationships between a subject (like a movie) in its triple store to a person (not necessarily an actor but maybe an historical figure), Dakick also can make the connection that a user that follows Franklin D. Roosevelt, for instance, would be interested in a documentary playing at a nearby theatre in the next week.
Build It Up
More, however, needs to be done to refine marking up data for an events-focused service. To that end, Dakick has been having conversations with Google and Yandex, the number one and two search engines in the country, about how schema.org could be even more helpful here.
As an example, because it is an events-focused domain, Dakick knows what it’s like to wrangle with issues such as whether 3D is a feature of a movie, an event of a movie, or a feature of a movie theater. How does that all play out for recommendations of World War Z to Brad Pitt followers around the different types of presentations they can choose from – 3D, Imax, and so on. “The movie is the same but the presentation is different,” he says. “The structure should be well-defined around that,” but comments must be applied to the movie itself, not to specific presentations of it.
Dakick also has had to expand or build from scratch data structures for other domains, like exhibitions and trade shows. It’s sharing the work it’s done with the search engine players to get their feedbacks on incorporating that into schema.org.
As Ünsal sees it, this all could help with Dakick potentially bringing more semantic context to search engine results, particularly with Google. With a Google search for Brad Pitt movies, you’ll enjoy being able to visually scroll through the sum of his movie career, clicking on any movie image to refine the search and get theatre showtimes, as well as enjoy tidbits from the Knowledge Graph related to his birth date, marriage, children and such. But Dakick thinks there’s opportunity to add to that information right in the search engine that Legends of the Fall, for instance, is playing on AMC that same night. “We can do that. To do that, there should be location-aware linked data with great content,” he says.
Or, if you search for Steve Jobs, neither will you get the information that there is a documentary about Steve Jobs tonight on NBC, he adds. “In Google the resulting information is static data, not context,” says Ünsal. “We may add context to them. We want to work with Google to surface more contextual information in its Knowledge Graph information.”
Dakick says it’s dedicated to having an open ecosystem, making all its semantic data open to the public. Its even, creative work, people and place ontologies are available here, and users can write SPARQL queries against them. In terms of building a profitable business, it sees the possibility to generate revenue in a few different ways, such as leading generation for ticket companies or online TV and video on demand services, or for music companies too. “Because of what we know we can measure the similarity of a singer with another band or another singer,” says Ünsal. “We know the data relations, the singer relations and band relations much better than companies in Turkey focused on music.”
The company is also testing the service in San Francisco.