Aspect Based Sentiment Analysis for Users Review Dataset Using Deep Learning and BERT

Sentiment Analysis is a crucial part in Natural Language Processing (NLP), Aiming to relate pre-defined labels/categories to a given text sentence or sequence. It is very well recognised not only in academia but also in the industry, giving real-time outputs via internet reviews on websites like Amazon, which can utilise the customer’s opinions on their products and services. The assumption of this task is that the entire text has an all-inclusive polarity. In this paper we aim to do sentiment analysis with BERT on the Review Dataset Collected by us. We did annotation of the data (11237 sentences) in the preprocessing phase which is one of the most crucial parts of the process then we use the outputs for the model as inputs that will be implemented on the same. We describe three classes related to the sentence idea namely Usefulness, Explanation and Competence and two classes for polarity namely Positive and negative. The output we get is the detailed sentiment analysis of the input review on the basis of the classes mentioned.