Dependency Tree-Based Contrastive Attention for Aspect-Based Sentiment Analysis

To address the issue of insufficient correlation between sentiment words and aspect words in sentiment analysis of text with complex syntactic structures, this paper proposes a novel sentiment analysis model based on a dependency tree-guided contrastive interactive attention mechanism—FtRoBERTa-GCN-TIAM. The model first fine-tunes RoBERTa to better capture syntactic structures, and optimizes word dependencies through Graph Convolutional Networks. Next, a contrastive interactive learning module is introduced, where a path-aware attention mechanism enhances the model's ability to capture the relationship between aspect words and sentiment words, thus reducing their distance and improving sentiment analysis accuracy. Experimental results show that the FtRoBERTa-GCN-TIAM model significantly outperforms traditional baseline models on multiple datasets, confirming the feasibility and effectiveness of the proposed model.