Automatic Identification of Zero Pronouns and their Antecedents within Aligned Sentence Pairs

要旨

This paper propose a method to identify sero pronouns within a 3apanese sentence and their antecedent equivalents within the corresponding English sentence from aed sentence pairm The method focuse on the characteristics of Japanese and English, in two lauguages from diffierent frnBies and in which distribution of zero pronouns is very different. In this method, the Japanese sentence and Engh ttausIatlon within the 3panese and English aligned sentence pairs are analyzed. Then, the pairs of $panese word/phrase and their English equivalent word/phrase are identified from each atigned sentence pair. Next, zero pronouns within a 3apaneze sentence are identified by using the syntactic and semantic structure of tYae Japane sentence and their antecedents within the English sentence are identified by using the characteristics of auaphoric and deictic expressions i English. This method was implemented using the Japaneze-to-English machine translation system, ALT-J/]] for the analysis of 3apaese entences and Bxin,s tagger for the analyis of the Enflish sentences. According to ny evaluation, for 554 zero pronouns i a sentence set for the evaluation of apanese-ta-Englih machine translation systems, 91.5% of the paLrs of zero pronun in tle Japanese sentences dud their mtecedents in the English tranalations were automatically identified correctly.

主題