Improving Span-based Aspect Sentiment Triplet Extraction with Abundant Syntax Knowledge
Aspect sentiment triplet extraction is a subtask of aspect based sentiment analysis, which has attracted considerable attention in recent years. It aims to obtain the aspect term, the associated opinion term and sentiment polarity, so called the triplet in reviews. Most existing studies can only learn the word-level context information and capture the superficial interactions between aspect and opinion terms. In contrast, our proposed span-based aspect sentiment triplet extraction model incorporates abundant syntax information, namely the span-level part-of-speech, constituent syntax, and dependency syntax information. Therefore, our model can learn both the connections between multiple words within the span and the interactions between aspect and opinion terms, and thus improve the performance on triplet extraction. Extensive experimental results show that our model outperforms previous models on four benchmark datasets and achieves state-of-the-art performance.
PDF journal 2022 PDF journal 2022 Abstract
