infotainment-orchestrator & recommender

The infotainment orchestrator takes care of the collection, indexing and retrieval of social content around large events. To a large extent, its main research components are described in the EventSense paper (Schinas et al., 2013) and in D2.2. The orchestrator coordinates the collection of content from Twitter, Facebook and other social media sources using as input the event hashtag, its Facebook page and event name. The orchestrator also arranges the operation of the following components (cf. figure below):

• Entity Extractor: For each incoming Item, the Entity Extractor detects references to entities of interest for the event, e.g. films. To achieve high-accuracy, it makes use of a bi-lingual entity gazetteer (available through the infotainmentDB).
• Sentiment Analyzer: Same as the one used by the news-orchestrator.
• Dysco Creator: The DySCO Creator clusters incoming Items based on the document-pivot method described in D2.2 and in (Aiello et al., 2013).
• Title Extractor: The title extractor is the same as the one used by the news orchestrator.
• Contributor Statistics: For each contributor (twitter account), some aggregate statistics are computed regarding their activity and the sentiment of their posts.
• Film Statistics: For each film, aggregate statistics are computed regarding their rating, popularity (based both on the myFilms feature of mobile app and tweets about the film), and sentiment.
• User Profiler: For each SocialSensor user (mobile app user), we collect the ratings and bookmarks (films added to myFilms). In case the user is logged in with their Twitter account, we also collect their followers and friends for use by the recommender.

In addition to the infotainment-orchestrator, the recommender, which implements the recommendation methods described in D5.2, runs as an independent service that makes use of the user profiles and ratings through the BigIndex server (cf. D4.2). The indexing of user profiles and ratings is handled by the orchestrator, while a REST service has been set up to serve the actual recommendations to the end users (of the mobile app).