Syntactic Approach Of Pronominal Anaphora Resolution In Intellectual Information Systems

Abstract

The algorithm, based on syntactic approach to pronominal anaphora resolution in patents in order to resolve anaphoric references and not to lose the information behind those, is described in this paper. The implemented system includes three modules. The first module filters non-anaphoric references, the second one attributes scores to the candidates based on the syntactic information extracted from tagged and parsed text and the third one chooses the candidate which obtained the maximum score. The system was tested on the corpus of US patents.