Synchronously tracking entities and relations in a syntax-aware parallel architecture for aspect-opinion pair extraction
Yue Zhang,Tao Peng,Ridong Han,Jiayu Han,Lin Yue,Lu Liu
DOI: https://doi.org/10.1007/s10489-022-03286-w
IF: 5.3
2022-03-12
Applied Intelligence
Abstract:Aspect-Opinion Pair Extraction (AOPE) task aims to capture each aspect with its corresponding opinions in user reviews. Entity recognition and relation detection are two fundamental subtasks of AOPE. Although recent works take interaction into account, the two subtasks are still relatively independent during calculation. Furthermore, since AOPE task has not been formally proposed for a long time, syntactic information does not attract much attention in the current deep learning models for AOPE. In this paper, we propose a model for Synchronously Tracking Entities and Relations (STER) to deal with AOPE. Specifically, we design a network consisting of a bank of gated RNNs, where we can track all entities of a review sentence in parallel. STER utilizes three features, i.e., context, syntax and relation, to learn the representation of each tracked entity and calculate the correlated degree between all entities synchronously at each time step. The entity representation and the correlated degree are highly dependent during calculation. Finally, they will be used for entity recognition and relation detection, respectively. Therefore, in STER, the two subtasks of AOPE can achieve sufficient interaction, which enhances their mutual heuristic effect heavily. To verify the effectiveness and adaptiveness of our model, we conduct experiments on two annotation versions of SemEval datasets. The results demonstrate that STER not only achieves advanced performances but adapts to different annotation strategies well.
computer science, artificial intelligence