Abstract
In recent years, sentiment analysis has appeared in our life to a great extent and has become a part of our life, but the previous coarse-grained sentiment analysis can no longer be satisfied with the current life[1] [2]. Aspect-Based Sentiment Analysis (ABSA) aims to identify aspect terms, and their corresponding sentiment polarities and perspectives. Therefore, ABSA has become a trend of life now[3]. However, the current pre-trained language models cannot perfectly solve the semantic information between sentences. In our paper, We propose a new way for IECSA to face the classification problem of ABSA, by using the CLS in Bert to learn better results, so that it can effectively learn more semantic information