Bi-RNN and Bi-LSTM Based Text Classification for Amazon Reviews

Abstract

Electronic Commerce (E-Commerce) enables the effective implementation of product-based online business transactions. In this paper, we categorize the dataset consisting of Amazon reviews into positive and negative. Consumers are able to choose a product solely on these binary text categorizations, neglecting the manual rating provided in the reviews. Recently, Deep Learning (DL) approaches started gaining popularity in E-Commerce applications. DL Techniques offers a powerful and effective approach for data analysis through filtering humongous E-Commerce data and finding hidden patterns and valuable details. Here, the research focuses on investing DL techniques for categorizing the text into positive and negative reviews. In this paper, we implemented two different DL approaches and found the accuracy of both approaches. DL models used here are bidirectional recurrent neural networks (Bi-RNN), and bidirectional long-short-term memory (Bi-LSTM). The accuracy obtained with Bi-RNN and Bi-LSTM is 89.60% and 92.80% respectively. The result shows that Bi-LSTM performs best on the Amazon review dataset.