New York University KBP 2010 Slot-Filling System

Document retrieval The entire text corpus was indexed using the Lucene document retrieval package1. As the first step in processing a query, Lucene was invoked to retrieve the documents containing the query name; a maximum of 500 documents were retrieved. No attempt was made to use the document associated with the name in the query in order to disambiguate possibly ambiguous names. As a minimal acknowledgement of possible name variation, if a person entity query name contained an initial or an organization entity query name contained a corporate suffix (“Corp.”, “Ltd.”, etc.) an alternate name was created omitting these items, and we retrieved documents with either the original or alternate name.