Enhanced Multi-Channel Graph Convolutional Network for Aspect Sentiment Triplet Extraction

Proceedings of the 60th Annual Meeting of the Association for Computational Linguistics Volume 1: Long Papers, pages 2974 - 2985 May 22-27, 2022 c⃝2022 Association for Computational Linguistics Enhanced Multi-Channel Graph Convolutional Network for Aspect Sentiment Triplet Extraction Hao Chen∗, Zepeng Zhai∗, Fangxiang Feng, Ruifan Li†, Xiaojie Wang School of Artificial Intelligence, Beijing University of Posts and Telecommunications, China {ccchenhao997, zepeng, fxfeng, rfli, xjwang}@bupt.edu.cn Abstract Aspect Sentiment Triplet Extraction (ASTE) is an emerging sentiment analysis task. Most of the existing studies focus on devising a new tagging scheme that enables the model to ex- tract the sentiment triplets in an end-to-end fashion. However, these methods ignore the relations between words for ASTE task. In this paper, we propose an Enhanced Multi-Channel Graph Convolutional Network model (EMC- GCN) to fully utilize the relations between words. Specifically, we first define ten types of relations for ASTE task, and then adopt a biaffine attention module to embed these rela- tions as an adjacent tensor between words in a sentence. After that, our EMC-GCN transforms the sentence into a multi-channel graph by treat- ing words and the relation adjacent tensor as nodes and edges, respectively. Thus, relation- aware node representations can be learnt. Fur- thermore, we consider diverse linguistic fea- tures to enhance our EMC-GCN model. Fi- nally, we design an effective refining strategy on EMC-GCN for word-pair representation re- finement, which considers the implicit results of aspect and opinion extraction when determin- ing whether word pairs match or not. Extensive experimental results on the benchmark datasets demonstrate that the effectiveness and robust- ness of our proposed model, which outperforms state-of-the-art methods significantly.1 1 Introduction Aspect Sentiment Triplet Extraction (ASTE) is a new variant of Aspect-based Sentiment Analysis (ABSA). The ASTE task aims to extract aspect sentiment triplets from a sentence, and each triplet contains three elements, namely aspect term, opin- ion term and their associated sentiment. In Figure 1, an example illustrates the definition of ASTE. *These authors contributed equally to this work. †Corresponding author. 1Code and datasets are available at https://github. com/CCChenhao997/EMCGCN-ASTE. Figure 1: A sentence with its dependency tree is given to illustrate ASTE task. In the triplet set, aspect terms, opinion terms are highlighted in blue and yellow, respec- tively. The positive sentiment polarity is highlighted in red, while the negative in green. To extract the triplets, previous studies have de- veloped three types of approaches. Pipeline ap- proaches (Peng et al., 2020) independently extract elements of the triplet. However, such techniques ignore the interaction between them, and poten- tially lead to error propagation and extra costs. To utilize the associations among the multiple sub- tasks, Mao et al. (2021) and Chen et al. (2021a) formulate the ASTE task as a multi-turn machine reading comprehension (MRC) problem and design a model based on BERT to jointly train multiple subtasks. Meanwhile, some efforts devote to ex- tracting the triplets in an end-to-end framework (Xu et al., 2020; Wu et al., 2020a; Zhang et al., 2020; Chen et al., 2021b; Yan et …

Content truncated to 500 words