Aspect-oriented Opinion Alignment Network for Aspect-Based Sentiment Classification

• DOI: 10.48550/arxiv.2308.11447
• Publication Date: 2023-01-01
• Authors: Xueyi Liu; Rui Hou; Yanglei Gan; Da Luo; Changlin Li; Xiaojun Shi; Qiao Liu
• Keywords: Sentiment Analysis; Benchmark (surveying); Polarity (international relations); Code (set theory)

Aspect-based sentiment classification is a crucial problem in fine-grained analysis, which aims to predict the polarity of given aspect according its context. Previous works have made remarkable progress leveraging attention mechanism extract opinion words for different aspects. However, persistent challenge effective management semantic mismatches, stem from mechanisms that fall short adequately aligning opinions with their corresponding multi-aspect sentences. To address this issue, we propose novel Aspect-oriented Opinion Alignment Network (AOAN) capture contextual association between and aspect. Specifically, first introduce neighboring span enhanced module highlights various compositions In addition, design multi-perspective align relevant information respect Extensive experiments on three benchmark datasets demonstrate our model achieves state-of-the-art results. The source code available at https://github.com/AONE-NLP/ABSA-AOAN.