Computational theory of short distance reflexive anaphoric devices in Urdu discourse for effective machine translation

Computational treatment of natural languages is a very challenging task. For effective machine translation of multi-sentential discourse units from a source language into some target language, resolution of anaphoric expressions is of crucial importance. This paper presents the computational treatment of the rarely studied form of pronominal reference called reflexives. A knowledge-poor, rule-based algorithm has been developed to identify and resolve the reflexives in Urdu discourses. The proposed algorithm resolves the reflexives in Urdu discourses with more than 85% accuracy.