SPIGA

Multilingual Personalized Press Reviews with Global Orientation

Competence CenterInformation Retrieval and Machine Learning
ContactProf. Dr.-Ing. Sahin AlbayrakDanuta Ploch
Partner: Neofonie GmbH

 

SPIGA creates multilingual, personalizable press reviews from a multitude of European and international news sources. SPIGA groups news according to real events, e.g. "CeBIT 2011", that are recognized automatically by the system. Instead of reading isolated news articles, users get a comprehensive overview of different news sources illustrating varying perspectives and the temporal development of news stories.

Semantic Topic Detection and Personalization

News articles are processed to identify, weight and disambiguate the main concepts and entities they contain. On the basis of this semantic profile, the system automatically recognizes if a new article corresponds to a known event, or constitutes a novel event (Topic Detection and Tracking). News articles are grouped into events using an incremental, centroid-based clustering approach. The semantic representations of events and news are used to guide the creation of personalized, topic-oriented press reviews that match concise selections of news articles satisfying a user’s information needs.

Together with our project partner Neofonie GmbH, DAI-Labor has developed a news aggregator that crawls documents from more than a thousand European news sources, and that uses the Apache UIMA Framework as the basis for a natural language processing pipeline that semantically annotates documents. The research focus of CC IRML's work lies on the development and evaluation of methods for Named Entity Disambiguation.

SPIGA is a cooperation project of DAI-Labor and Neofonie GmbH and is funded by the Federal Ministry of Economics and Technology.