Semantic Role Based Tamil Sentence Generator

Abstract

A Machine learning technique called memory based bigram models are developed for a system that generate a simple sentence for Tamil language from a set of concept terms and their semantic role. This system consists of a learner to learn how to realize a sentence from the content of semantic role information. This learner has been designed as a statistical model that is formulated from a preprocessed corpus of sentences. This preprocessing work is handled by annotating the corpus using part of speech tagging, chunking and semantic role labeling processes. This collective annotated corpus is statistically analyzed and developed the memory based bigram models. These models thus obtained are capable of producing the appropriate sequence of semantic roles of the concept terms for realizing sentence. A phrase generator has been developed to generate the appropriate phrases involved in sentence generation.