Sentiment analysis of online Tamil contents using recursive neural network models approach for Tamil language

This paper proposes an approach involving Recursive Neural Network models for Tamil in improving the accuracy of a sentiment analyzer tool for Tamil language. To capture the meaning of phrases and thereby detecting the sentiment of them, Naïve Bayes approach and Hidden Markov Model algorithm take into consideration frequency of occurrence of keywords. Vector space models use term-document, word-context and pair-pattern matrices for the same. None of these models can capture the meaning of long phrases, ambiguous or sarcastic phrases accurately. Sentiment detection requires richer resources than just frequency of keywords. Recursive Neural Network models can be generated for Tamil language based on which sentiment analysis can be done. The accuracy of the results obtained can be enhanced using inter sentential sentiment prediction.