A European Union-funded project to bring the web and TV closer together, called NoTube, wrapped up earlier this year. But its legacy lives on in the form of Beancounter.io from Sourcesense. The company, which was one of the project’s co-founders, had a role in the NoTube effort around integrating viewers’ social web activities as part of the platform to deliver TV content in personalized ways to users.
Leveraging the open source software, libraries and best practices that were outcomes of the project, Sourcesense has continued to move forward to deliver a commercial, scalable Web API platform that offers semantically enriched user profiles built from users’ activities performed on the Social Web. One of the first customers of its efforts is one of the largest Italian broadcast companies, RAI, which was also involved in the EU project.
Beancounter is powering a second- screen service on top of the platform to provide its 5 million viewers information on related content that may be of interest to them based on profiling their social activities (with their permission).
Beancounter.io works by using Linked Open Data identifiers to describe users’ interests represented on social networks Facebook, Twitter, Last.FM and Foursquare, says Sourcesense R&D director Davide Palmisano, who will be presenting a talk about the technology at next week’s Semantic Technology and Business Conference in the U.K.. “When we profile users in terms of interest or behavior, we represent interests like DBpedia identifiers. This makes our user profiles completely inter-operable, and useable by other applications. They can just access our profiles to see which DBpedia URIs are in there.”
As Palmisano explains linking users’ interest to DBpedia identifiers, the process starts using a mix of NLP techniques to extract identities like names of cities, politicians and so on. “If something is identified on DBpedia, then we grab it and link to it,” he says. The NLP techniques are not always useable, as in the case of users’ listening to a song on LastFM, he says, in which case MusicBrainz provides the identifiers that then link the song, artist and so on to DBpedia. Sourcesense has all the metadata that is indexed using the same techniques in order to make matches between interests and existing content. “Our profiles are updated in real-time,” Palmisano notes, “so the actions you do on the social web has immediate impact on your profile.”
Customers don’t have to rely on the DBpedia connection if they prefer to use their own databases, as RAI does, since interoperability isn’t necessary for its implementation of recommending its own content to users. They also can plug in their own choice of NLP technologies to fit their domains best, he says.
While TV broadcasters are one natural fit, Palmisano sees applications beyond that – in, for example, ecommerce. “It’s good for delivering customized products to customers or to provide data to advertisers, because all the profiles we compute and syndicate are stored and we can perform match analysis on top to see related interests.”
Sourcesense is working on creating batch analytics offerings it can sell as services to customers. There’s still time to catch them at SemTech, which you can register for here.
- Big Data Skills Worth Big Bucks
- Automatic Hashtags & Machine Learning: The New Google+
- Gmail, Meet JSON-LD
- Cambridge Semantics Wins SIIA Software CODiE Award