Beyond Traditional Methods: The Power of Bi-LSTM in Transforming Customer Review Sentiment Analysis

Abstract

In the current generation, many large and small companies compete fiercely to create things better than those on the market, such as smartphones, TVs, and many other things. One way they can do this is by guaranteeing the quality of services or goods that are better than others. The provider must investigate the feedback of their users or customers to improve the quality of service of the goods or services offered. Most medium and small companies, such as Micro, Small, and Medium enterprises (MSMEs), online stores, and so on, conduct research on customer feedback manually by looking at one-by-one feedback from customers, which is very ineffective and inefficient if a lot of customer feedback is obtained. Therefore, this research is conducted with the intention and purpose of helping medium and small companies analyze their customer sentiment, as well as trends over a certain period. This research will apply the Bidirectional Long Short-Term Memory (Bi-LSTM) algorithm to perform sentiment analysis on customer feedback. This research also compares other deep learning methods with the proposed method, namely the Uni-LSTM, GRU, CNN, and Simple-RNN algorithms. After testing, the accuracy results of the Uni-LSTM, Bi-LSTM, GRU, CNN, and Simple-RNN algorithms are 52.2%, 92.4%, 52.2%,