Heuristic Algorithm for Resolving Pronominal Anaphora in Hindi Dialects

Abstract

Artificial intelligence is a necessity for today’s realistic world knowledge. The facts hidden by the anaphoric expressions can be revealed by anaphora resolution only. The relevance of anaphora resolution could not be avoided as its productivity affects the performance of text summarization, automatic question answering system, information extraction, etc. The paper comes up with implementations of algorithms for first, second, and third-person pronouns for Hindi language, with the ability to tackle intersentential anaphora within the scope of 3–5 sentences and subsequently the algorithms can be tailored-up for more sentences. Approximately, 698 sentences were experimented for feature selection of each type of personal pronoun. The proposed algorithms have been tested on the synthetic datasets of 1059 sentences which contain 712 pronominal anaphors out of 781 anaphoric pronouns. The F-measure evaluation for selected corpora gives promising results, indicating that the algorithms are effective in resolving Hindi pronominal anaphora.