A Real-Time Machine Learning Approach for Sentiment Analysis

Abstract

The rapid increase in data volume and human concern about quick information mount the need for knowledge discovery with meager time span. Discover facts (data) about real entity and derive conclusions (information) from those facts and storing them for future use and reference (knowledge), is an art to get the true feeling and sentiment. In recent trend, knowledge discovery and knowledge management has been highly influenced by Sentiment Analysis. Sentiment Analysis provides contextual polarity of a document with respect to some issues or some topic .This paper contributes a new approach of deploying Artificial Neural Network and k-Mean algorithm for Sentiment Analysis. The approach incorporate the linguistic analysis of different components of a sentence(Adverb, Adjective, Noun, Verb) into a artificial neural network for supervised learning and k-Mean algorithm for unsupervised learning and the output (five cluster representing strong like, weak like, doubtful, weak dislike and strong dislike) from the network will not only simplify e-Discovery(rapid identification of potentially relevant data) solutions and opinion analysis system but also shown significant advancement from the previous research on this domain. The system not only classify documents and provide a relevant information but also optimizes steps of different techniques used for Sentiment Analysis and increases the performance (reducing memory and processor utilization) by modifying the deployed algorithms.