Fine-Grained Sentiment Analysis Based on Heterogeneous Graph Neural Network
Xueru Bai, Rong Fei, Zuo Liu, Xiongbo Chen · 2022
Fine-grained sentiment analysis is a key task of sentiment analysis that aims to identify emotional polarity in a particular aspect of a sentence. Existing research has shown that dependency information can improve the performance of sentiment analysis task. In recent years, many studies have begun to focus on global dependence information, and these studies have also significantly improved the performance of emotion analysis tasks. In this paper, we present the GLo-GCN method. Specifically, it uses the results of dependent analysis to learn local dependence information weight, on the whole corpus to build words and document heterogeneous graph, and the heterogeneous graph neural network further divided into more fine-grained graph neural network learning global dependent information weight, eventually fusion local dependence information weight and global dependence information weight for emotional analysis. Performance on two public datasets demonstrates the effectiveness of our model over state-of-theart models.