Computational Approach for Improving Fluency in Bangla NLG

The primary goal of Natural Language Generation (NLG) is to generate complete and fluent natural language text from underlying machine representation of information. The completeness of a text depends on its semantic correctness and quality of being understood. The fluency of a text depends on its grammatical correctness, coherence and naturalness.The fluency of a text generated is as important as its completeness. In NLG, prenominal modifier ordering and syntactic aggregation are two useful methods for improving the generated text fluency. Correct ordering of prenominal modifiers, modifying the same head noun, improves the fluency and naturalness of text. In syntactic aggregation simple text spans are combined by using linguistic rules. Unnecessary repeating words are deleted by this process thereby improving the fluency, coherence and conciseness of the text. In this thesis, we have presented computational approaches for these two problems in Bangla NLG.