Research on Multi-class Sentiment Analysis of Social Media Texts Based on the ERNIE Model

Abstract

With the rapid development of social media, how to accurately gain insights into the public's sentiment tendencies from vast amounts of user text data has become an urgent problem to be solved. To effectively address this challenge, this research proposes an emotion analysis model named ERNIE, which focuses on multi-class sentiment analysis on social media. The construction of this model aims to solve the problem that existing sentiment analysis technologies have difficulty in precisely identifying complex sentiment categories. By manually annotating five-class texts, the ERNIE model retains the multi-level semantic information of the texts, thereby being able to more accurately capture the sentiment details in online comments. Using a large amount of comment data, the ERNIE model conducts an in-depth analysis of these data and summarizes the key information contained in the online comments. Through performance comparisons with different algorithms, the ERNIE model exhibits excellent performance in key indicators such as accuracy, recall, and F1 value, and finally determines the optimal model combination. The research results show that the ERNIE model has achieved remarkable results in the multi-class sentiment analysis of social media texts, being able to accurately distinguish sentiment categories such as very positive, positive, neutral, negative, and very negative, providing strong technical support and a basis for decision-making in fields such as public opinion monitoring, user opinion mining, and market trend analysis.