Sentiment Analysis on Bangladeshi E-commerce Product Review Data

Abstract

In this paper, we propose a sentiment analysis process for online product reviews written in Bengali. In recent years, with the rapid development of Internet technology, online shopping has become a dominant method of user purchasing and consumption. Analyzing the emotions of multiple numbers of user reviews on e-commerce platforms can help determine whether a user's attitude is positive, neutral or negative, it captures the opinions, beliefs and feelings of each user about the respective product. This work presents a machine learning-based technique to identify sentiment polarities (positive, negative or neutral category) from online product reviews in Bengali. To evaluate the effectiveness of the proposed method, a corpus containing 2228 reviews was extracted using the Daraz API of Bengali online product reviews is being developed. Comparative analysis with various approaches LR, DT, RF, MNB, KNN, SVM, SGD etc. Consider the characteristics of unigram, bigram, and trigram features, respectively. Experimental results reveal that the stochastic gradient descent (SGD) with the Tri-gram feature outperforms the other techniques with 88% accuracy.