Complex Pronoun Resolution using Large-scale Dependency Knowledge
Masaki Onishi · Institutional Repositories DataBase (IRDB) · 2016
Coreference resolution is to cluster all text spans which refer to the same entity in a document.Pronoun resolution is one of most difficult task of coreference resolution.It is well kwnon that pronouns are easily disambiguated by human but difficult for a computer, and not solvable by simple linguistic techniques.The Winograd schema is a pair of twin ambiguity sentences that include little difference and to anaphora another entity created by Levesque.In this paper, I use instance-based coreference resolution.I collect many sentences including coreference from Gigaword and use those depenpency subtrees as knowledge.This paper show that the existing sentences' knowledge of dependency subtrees is useful for difficult pronoun resolution.