Automatic Classification of Hindi Verbs in Syntactic Perspective

We report of a rule based, knowledge-base driven tool to automatically classify Hindi verbs in syntactic perspective. We also report of developing the largest lexical resource for Hindi verbs along with the information on their class based on valency and some syntactic diagnostic tests as well as their morphological/inflectional type. We use this resource to develop the tool to automatically classify Hindi verbs in any given input text. In this process we also develop an algorithm based on the analysis of this lexical resource and a large Hindi corpus. In this algorithm we spell out the rules to disambiguate the homophonous/homo-gramphaemic words that become the root cause of ambiguity in this task. The tool developed has been tested against two large annotated corpora of Hindi and achieves close to 100% result on a qualitative evaluation and more than 99% on quantitative evaluation if errors from other sources are discounted. Keywords—Automatic Verb Class Identification, Hindi, NLP, Verb Class Knowledge Base, Verb Class Disambiguation