The main aim of the project was the development of a television contents recommender based on a novel automatic categorization approach. The project considered a very large set of un-structured and semi-structured data (blogs, news web sites, newspapers) as user context, and enriched the existing categories depending on the meanings assumed by them in the most recent contents. Therefore, the categories were organized in structured taxonomies depending on the semantics expressed by the considered domains. These taxonomies have been finally used for indexing the television contents by using novel classification algorithms; using these associations, the system was able to highlight hidden semantic relationships among indexed TV contents, defining the basis for an intelligent television contents recommender. The results of this project have been patented (Pub. No.: WO/2010/075889).

The CADMOS project aims at designing a novel semantic annotation model for multimedia documents (video, text, audio) that takes into account their narrative features (characters involved in the narrated story, their actions, goals, etc). The project consists in devising a BDI annotation model that also integrates vast commonsense ontological knowledge and a novel ontology- based model of narrative, Drammar, to permit the annotation of the narrative features of the considered documents. The project also aims to develop a web-based software platform to annotate and access the archived documents.