Ensemble Sentiment Analysis Using Bi-LSTM and CNN
Puneet Singh Lamba, Achin Jain, Harsh Taneja, Prakhar Priyadarshi, Arvind Panwar, Arun Kumar Dubey
Abstract
Sentiment analysis is an invaluable skill for categorizing or evaluating points of view of people from all over the world. Understanding how people feel about a certain issue, product, or service is helpful. Sentiment analysis has been a focus of several studies, with most methods based on natural language processing techniques. Nevertheless, determining sarcasm, negation, or ambiguity in language usage has always proven challenging. In this research, we present an ensemble approach that includes methods such as Keras Sequential CNN and Bi-Long Short-Term Memory (Bi-LSTM). Two different datasets, the YELP review dataset and the IMDB review dataset, were used to evaluate the model. The outcomes were good, and our model performs well in comparison to a few other methods used to a related problem. Even though Bi-LSTM and Keras Sequential CNN produced good results on their own, employing ensemble techniques improved them even more.