Machine Learning Approach for Resolving Pronominal Anaphora Using Hindi Dependency Treebank

Machine Learning facilitates the computers to mimic human intelligence by applying a set of rules to massive amounts of trained data and identifying patterns to make decisions and adapt based on what patterns are still uncovered. A number of applications ranging from spam detection, facial recognition, product recommendations to credit-card fraud detection, all of them apply machine learning procedures. The focus is on presenting machine learning approach for resolving anaphora’s in Hindi Sentences. The availability of Dependency Treebank for Hindi has motivated many researchers to explore and exploit its information for natural language processing such as anaphora resolution. Capturing the Treebank generated by a parser has been seen as a key element in resolving anaphora. An attempt is to show how the part-of-speech (POS) tagging, chunking and morphological information generated by the Hindi parser in the form of Hindi Dependency Treebank (henceforth HDT) can be used to derive rules for resolving Hindi anaphora and implementing the same in machine learning. The steps for resolution of pronominal anaphora are based on the syntactic cue provided by the HDT.