A Pairing Enhancement Approach for Aspect Sentiment Triplet Extraction
This work improves ASTE for sentiment analysis in review texts, but it is incremental as it builds on existing methods with a novel training enhancement.
The paper tackled the problem of Aspect Sentiment Triplet Extraction (ASTE) by addressing confusion in connecting aspect and opinion terms, proposing a pairing enhancement approach with contrastive learning that performs well on four datasets compared to classical and state-of-the-art methods.
Aspect Sentiment Triplet Extraction (ASTE) aims to extract the triplet of an aspect term, an opinion term, and their corresponding sentiment polarity from the review texts. Due to the complexity of language and the existence of multiple aspect terms and opinion terms in a single sentence, current models often confuse the connections between an aspect term and the opinion term describing it. To address this issue, we propose a pairing enhancement approach for ASTE, which incorporates contrastive learning during the training stage to inject aspect-opinion pairing knowledge into the triplet extraction model. Experimental results demonstrate that our approach performs well on four ASTE datasets (i.e., 14lap, 14res, 15res and 16res) compared to several related classical and state-of-the-art triplet extraction methods. Moreover, ablation studies conduct an analysis and verify the advantage of contrastive learning over other pairing enhancement approaches.