NiCT at TREC 2010: Related Entity Finding.
Youzheng Wu, Chiori Hori, Hisashi Kawai · Text REtrieval Conference · 2010
This paper describes experiments carried out at NiCT for the TREC 2010 Entity track. Our studies mainly focus on improving the NE Extraction and Ranking Entity modules, both of them play vital roles in Related Entity Finding system. In our last year’s system, only a Named Entity Recognition tool is used to extract entities that match coarse-grained types of target entities such as organization, person, etc. In this year, dependency tree-based patterns learnt automatically are employed to filter out entities that do not match fine-grained types of target entities such as university, airline, author, etc. In the Entity Ranking part, we propose a dependency tree-based similarity method and incorporate homepage information to improve ranking.