BERT Model with Fuzzy Logic Optimization on Multivariate Sentiment Analysis Tasks

Abstract

With the development of natural language processing techniques, sentiment analysis techniques are constantly updated. However, most of the sentiment analysis techniques are limited to handling binary sentiment analysis and perform poorly on multivariate sentiment analysis tasks. This paper explores the optimization in the field of sentiment analysis by using the BERT model combined with fuzzy logic to improve the performance of the model on multivariate sentiment classification tasks. The study uses the Twitter comments dataset and aims to gain insights into the challenges and optimization strategies of sentiment analysis models when dealing with multivariate sentiment. First, a single BERT model is experimentally demonstrated to perform better in binary classification tasks compared to multivariate classification tasks, highlighting its adaptability for simple sentiment classification and providing a basis for comparative experiments for multivariate classification experiments. To further improve the performance in multivariate sentiment analysis, fuzzy logic is introduced in this paper, and through the integration of this fuzzy logic, the accuracy of the BERT model on the multivariate sentiment classification task is successfully improved. The results of the study not only reveal the superiority of a single BERT model in binary classification but also emphasize the importance of incorporating fuzzy logic when facing multivariate sentiment analysis.