Sentiment Analysis and Topic Modeling on News Headlines

Abstract

Sentiment analysis and topic modeling has wide range of applications from medical to entertainment industry, corporates, politics and so on. News media play vital role in shaping the views of public towards any product or people. The dataset used for this work is news headlines dataset of one of the leading new portals of India i.e., Times of India. This research aims to perform comparative study of both supervised and unsupervised learning for text analysis and use the best performing models in both the category for prediction of sentiment and topic classification of news headlines. For sentiment analysis, supervised techniques like Machine learning ensemble model and Bi-LSTM have used. Similarly, unsupervised techniques like LDA (Latent Dirichlet Allocation) and LSA (Latent Semantic Analysis) have been for topic modeling.