A fresh look on knowledge bases: Distilling named events from news

Knowledge bases capture millions of entities such as people, companies or movies. However, their knowledge of named events like sports finals, political scandals, or natural dis-asters is fairly limited, as these are continuously emerg-ing entities. This paper presents a method for extract-ing named events from news articles, reconciling them into canonicalized representation, and organizing them into fine-grained semantic classes to populate a knowledge base. Our method captures similarity measures among news articles in a multi-view attributed graph, considering textual con-tents, entity occurrences, and temporal ordering. For dis-tilling canonicalized events from this raw data, we present a novel graph coarsening algorithm based on the information-theoretic principle of minimum description length. The qual-ity of our method is experimentally demonstrated by ex-tracting, organizing, and evaluating 25 000 events from a corpus of 300 000 heterogeneous news articles