STAR：基于关系学习的分步任务增强用于方面情感四元组预测
摘要
方面基于情感分析（ABSA）旨在识别四个情感元素，包括aspect term、aspect category、opinion term和sentiment polarity。这些元素构成了情感的完整图景。最具挑战性的任务是aspect sentiment quad prediction（ASQP），它需要同时预测这四个元素，并且受难以准确建模情感元素之间依赖关系的限制。一个关键挑战在于注释数据的稀缺，这限制了模型理解和推理所需的关系依赖以实现有效的quad预测的能力。为解决这一挑战，我们提出了一个基于关系学习的分步任务增强框架，将ASQP分解为一系列具有递增关系粒度的辅助子任务。 Specifically, STAR incrementally constructs auxiliary data by augmenting the training data with pairwise and overall relation tasks, enabling the model to capture and compose sentiment dependencies in a stepwise manner. This stepwise formulation provides effective relational learning signals that enhance quad prediction performance, particularly in low-resource scenarios. Extensive experiments across four benchmark datasets demonstrate that STAR consistently outperforms existing methods, achieving average F1 improvements of over under low-resource conditions.
引用
@article{arxiv.2501.16093,
title = {STAR: Stepwise Task Augmentation with Relation Learning for Aspect Sentiment Quad Prediction},
author = {Wenna Lai and Haoran Xie and Guandong Xu and Qing Li},
journal= {arXiv preprint arXiv:2501.16093},
year = {2026}
}
备注
17 pages, 6 figures, and 7 tables