Artificial intelligence

Abstract: The article is dedicated to modern algorithm of pronominal anaphora resolution. Anaphora resolution should be considered in a wider range of problems related with language ambiguity resolution, for instance: entity recognition, reference analysis and in general case, of course, semantic analysis of natural language text. We can render conclusion from stated above that anaphora resolution is possible only on semantic level of natural language analysis. The main purpose of this work is development of semantic heuristics for finding the most probable antecedent corresponding to anaphora with analysis of sentence context. The proposed algorithm gives about 5% improvements in comparison to the standard Mitkov algorithm.

Keywords: natural language text processing, anaphora resolution, semantic analysis.

References: