Opinion Mining in Tamil Tweets using Dense Neural Network (DNN) Architecture
Abstract
Tamil is one of the Indian languages which still requires the state of the art model for Opining mining or Sentiment Analysis(SA). It aims to extract the opinions from tweets, reviews and comments for development of NLP(Natural Language Processing) in Tamil Language. Therefore, this article proposed a Neural Network model to analyze the Tamil tweets into two polarities as Positive and Negative. The work comprises different phases such as its own Tamil corpus creation with 1015 Tamil tweets, Word Vectorization and Classification. A Deep learning based Dense Neural Network (DNN) architecture was trained with word vectors for finding the positive and negative comments. The proposed model works well in the manually annotated corpus without removing the stop words and sentimental symbols like emojis’. It is observed that the proposed model gives better accuracy of 86% when compared with the Machine Learning algorithm.