Sentiment Analysis on a Multi-Class Binary Sentiment Dataset Using Bi-LSTMs and BERT
DescriptionUnderstanding customer sentiments towards a product has become of paramount importance to many business sectors. Thanks to the availability of large amount of customer reviews, companies can now have valuable insights on how customers perceive their product and hence design ways to improve on their products and services. The growth of social media such as Twitter, social networks and blog posts has led to growing importance of sentiment analysis, thus making it one of the most active research areas in the Natural language processing (NLP) Community. Sentiment Analysis is a common use case in NLP where sentiments are classified as Positive, Negative and Neutral depending on the customers product review
In this paper, we perform sentiment analysis on a multi-class binary sentiment dataset of customer reviews using two NLP algorithms - the Bi-directional Long Short-Term Memory (Bi-LSTM) and the Bidirectional Encoder Representations for Transformers (BERT). Then, we will compare their performance to determine which is best suited for our datasets. Our goal for creating this multi-class dataset is to help us explore how feasible it is to use a single model on three different datasets at the same time.