Sentiment Analysis using Telugu SentiWordNet

Abstract

In recent times, sentiment analysis in low resourced languages and regional languages has become emerging areas in natural language processing. Researchers have shown greater interest towards analyzing sentiment in Indian languages such as Hindi, Telugu, Tamil, Bengali, Malayalam, etc. In best of our knowledge, microscopic work has been reported till date towards Indian languages due to lack of annotated data set. In this paper, we proposed a two-phase sentiment analysis for Telugu news sentences using Telugu SentiWordNet. Initially, it identifies subjectivity classification where sentences are classified as subjective or objective. Objective sentences are treated as neutral sentiment as they don’t carry any sentiment value. Next, Sentiment Classification has been done where the subjective sentences are further classified into positive and negative sentences. With the existing Telugu SentiWordNet, our proposed system attains an accuracy of 74% and 81% for subjectivity and sentiment classification respectively.