STAR: Stepwise Task Augmentation and Relation Learning for Aspect Sentiment Quad Prediction
This work addresses data scarcity in aspect-based sentiment analysis, an incremental improvement for natural language processing applications.
The paper tackles the challenge of insufficient annotated data for aspect sentiment quad prediction (ASQP) by proposing STAR, a strategy that constructs auxiliary data through stepwise task augmentation and relation learning, resulting in superior performance on four benchmark datasets.
Aspect-based sentiment analysis (ABSA) aims to identify four sentiment elements, including aspect term, aspect category, opinion term, and sentiment polarity. These elements construct the complete picture of sentiments. The most challenging task, aspect sentiment quad prediction (ASQP), predicts these elements simultaneously, hindered by difficulties in accurately coupling different sentiment elements. A key challenge is insufficient annotated data that limits the capability of models in semantic understanding and reasoning about quad prediction. To address this, we propose stepwise task augmentation and relation learning (STAR), a strategy inspired by human reasoning. STAR constructs auxiliary data to learn quadruple relationships incrementally by augmenting with pairwise and overall relation tasks derived from training data. By encouraging the model to infer causal relationships among sentiment elements without requiring additional annotations, STAR effectively enhances quad prediction. Extensive experiments demonstrate the proposed STAR exhibits superior performance on four benchmark datasets.