Rule Based Anaphora Resolution for ISL Generation System

Abstract

Sign language (SL) is the most well-known approach of articulation of thoughts and feelings for the deaf community. For transformation with one dialect then onto the next one needs itemized learning about the two dialects. There are relatively few specialists of gesture based communication. Machine interpretation of gesture based communication is an extreme assignment because of absence of phonetically commented on and all around recorded information on Indian Sign Language because of which inquire about on Indian Sign Language etymology and phonology is restricted. Sign language translation is used to convert a simple sentence into Indian Sign Language. While doing so a number of problems have been faced. Among them, Anaphora resolution is one such problem that comes across while converting sentence into ISL. Anaphora is a discourse level phenomenon whereby the understanding of an event of one articulation relies upon the translation of an event of another. Anaphora occurs frequently in written texts and spoken languages. The focus of this thesis is Entity Anaphora Resolution and pronominal forms. A web based system has been developed that works for Entity Anaphora using dependency information in Hindi. In this thesis, rule based approach which characterizes rules utilizing dependency structures and relations for referent determination have been developed. Rule based approach is the least difficult approach to take care of the issue of anaphora determination. Different tenets for various sorts of pronouns have been examined in this theory. This thesis discusses about the rule set and algorithms define and tested over a corpus of 300 sentences. Each pronominal form has an algorithm to be followed to find the referent corresponding to its anaphora. The algorithms are run against the type of anaphor to resolve it. The system produces the accuracy of 60%. The system resolves the anaphor for spatial pronouns more efficiently as compared to others. The system fails to intact the grammatically and syntactically meaning of the sentence.