Another Evaluation of Anaphora Resolution Algorithms and a Comparison with GETARUNS' Knowledge Rich Approach

1 Abstract In this paper we will present an evaluation of current state-of-the-art algorithms for Anaphora Resolution based on a segment of Susanne corpus (itself a portion of Brown Corpus), a much more comparable text type to what is usually required at an international level for such application domains as Question/Answering, Information Extraction, Text Understanding, Language Learning. The portion of text chosen has an adequate size which lends itself to significant statistical measurements: it is portion A, counting 35,000 tokens and some 1000 third person pronominal expressions. The algorithms will then be compared to our system, GETARUNS, which incorporates an AR algorithm at the end of a pipeline of interconnected modules that instantiate standard architectures for NLP. F- measure values reached by our system are significantly higher (75%) than the other ones.