"What's in the news?“ - or: why Angela Merkel is not significant

06/02/2015 - 14:00 to 14:20
short talk (20 min)

Session abstract: 

Bluekiwi.de, Unister's meta search engine, offers an explicit search over news articles. But what about the user who wants to inform herself on what is actually new and important? Watching hundreds of news sources, we collect thousands of news every day - far too many to read or present concisely. So the question becomes how to aggregate, summarize and present the important news of the day.

In this talk, we want to present some strategies we used to aggregate news, figure out what is important and not just common, and what techniques can be used to come up with descriptive teasers. These teasers could be used in turn to trigger useful search queries, thus providing the user with the full news stories behind them. Our approach comprises proper entity detection and natural language processing, which we will compare to pure term based techniques. Even if this approach is certainly more laborious than 'out-of-the-box' information retrieval, it is well worth the effort, providing us with understandable and succinct summaries of what's really going on.