An Adam Algorithm-Based Pretraining Model for Public Opinion Early Warning

With the rapid rise of short video (SV) platforms, the sentiment information contained in SV comments has become increasingly important to enterprises, governments, and other organizations. This paper focuses on researching sentiment analysis of SV comments and employs the pretraining model NEZHA in BERT for experimental validation. First, the SV comments are preprocessed, and feature extraction is performed. Then, using the NEZHA model, comment sentiment analysis is divided into two main categories: positive and negative. The experimental results demonstrate that the NEZHA model achieves a high level of accuracy in the field of SV comment sentiment analysis and has significant advantages over other methods. This study offers an efficient technical solution for analyzing the sentiment of brief video comments, aiding in the precise comprehension and response to public opinion.