SENTIMENT ANALYSIS ON MOBILE REVIEWS BY USING DEEP LEARNING MODELS

Abstract

The last two decades has been people posting about different aspects of their lives online. Be it their private life, their pets, parents, siblings, kids, etc. It has become a norm in the day-to-day life of different bloggers and people who post videos online. Opinion mining/Sentiment Analysis falls under Natural Language Processing and is used to judge the sentiment expressed by the people. The sentiment could be positive, negative or neutral. We have used deep learning techniques like Recurrent Neural Networks, Convolution Neural Networks amongst others to find sentiments of different users in the domain of mobile reviews. Bi-directional Encoder Representations from Transformers gives the best accuracy score of 98% and an f-score of 97%.