Multilingual Anaphora Resolution

This paper presents amultilingual robust, knowledge-poor approach to resolvingpronouns in technical manuals. This approach is a modification of the practicalapproach (Mitkov 1998a) and operates on texts pre-processed by apart-of-speech tagger. Input is checked against agreementand a number of antecedent indicators. Candidates are assigned scores by eachindicator and the candidate with the highest aggregate score isreturned as the antecedent. We propose this approach as aplatform for multilingual pronoun resolution. The robust approach was initiallydeveloped and tested for English, but we have also adaptedand tested it for Polish and Arabic. For bothlanguages, we found that adaptation required minimummodification and that further, even if used unmodified, the approachdelivers acceptable success rates. Preliminary evaluation reports high successrates in the range of over 90%.