SENTIMENT ANALYSIS OF ONLINE PRODUCT REVIEWS USING BERT

Computer Science & Engineering

The utilization of the BERT neural network model in evaluating emotions from online product reviews has demonstrated its effectiveness in enhancing the comprehension of customer preferences on online products. This approach not only aids platforms in better addressing customer needs but also assists Customers in identifying suitable and budget-friendly Products. This advancement contributes to the refinement of product recommendations through intelligent processing. By harnessing the pretrained BERT model, a series of experiments were conducted involving sentiment analysis. Through meticulous parameter adjustments throughout the experimentation phase, a highly accurate classification model was developed. The BERT layer was employed as a foundational word vector layer. Input text sequences were fed into this layer to undergo vector transformation. The resulting vectors, upon passing through corresponding neural networks, underwent classification via the SoftMax activation function. Sentiment analysis involves assessing the attitudes of customers based on textual data, particularly product reviews in this case. Through this study, we developed a solution capable of determining sentiments expressed in online product reviews. To achieve this, we employed algorithms such as Logistic Regression, Random Forest Classifier, and Sentiment Intensity Analyzer. The experimental outcomes illustrate the achieved accuracy in performing sentiment analysis.

This work is licensed under a Creative Commons Attribution-ShareAlike 4.0 International License.