Relation-Aware Collaborative Learning for Unified Aspect-Based Sentiment Analysis

Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics, pages 3685–3694 July 5 - 10, 2020. c⃝2020 Association for Computational Linguistics 3685 Relation-Aware Collaborative Learning for Unified Aspect-Based Sentiment Analysis Zhuang Chen, Tieyun Qian∗ School of Computer Science, Wuhan University, China {zhchen18, qty}@whu.edu.cn Abstract Aspect-based sentiment analysis (ABSA) in- volves three subtasks, i.e., aspect term extrac- tion, opinion term extraction, and aspect-level sentiment classification. Most existing studies focused on one of these subtasks only. Several recent researches made successful attempts to solve the complete ABSA problem with a uni- fied framework. However, the interactive re- lations among three subtasks are still under- exploited. We argue that such relations en- code collaborative signals between different subtasks. For example, when the opinion term is “delicious”, the aspect term must be “food” rather than “place”. In order to fully ex- ploit these relations, we propose a Relation- Aware Collaborative Learning (RACL) frame- work which allows the subtasks to work co- ordinately via the multi-task learning and re- lation propagation mechanisms in a stacked multi-layer network. Extensive experiments on three real-world datasets demonstrate that RACL significantly outperforms the state-of- the-art methods for the complete ABSA task. 1 Introduction Aspect-based sentiment analysis (ABSA) is a fine- grained task which aims to summarize the opin- ions of users towards specific aspects in a sentence. ABSA normally involves three subtasks, namely aspect term extraction (AE), opinion term extrac- tion (OE), and aspect-level sentiment classifica- tion (SC). For example, given a review “The place is small and cramped but the food is delicious.”, AE aims to extract a set of aspect terms {“place”, “food”}. OE aims to extract a set of opinion terms {“small”, “cramped”, “delicious”}. Meanwhile, it is expected for SC to assign a sentiment polarity “negative” and “positive” to the aspect “place” and “food”, respectively. *Corresponding author. R1 R4 R2 R3 AE OE SC ABSA – Aspect-Based Sentiment Analysis AE – Aspect Term Extraction OE – Opinion Term Extraction SC – Aspect-level Sentiment Classification Figure 1: Interactive relations among subtasks in ABSA (left), and a list of abbreviations (right). Most existing works treat ABSA as a two-step task containing AE and SC. They develop one sepa- rate method for each subtask (Tang et al., 2016; Xu et al., 2018; Li et al., 2018a; Hu et al., 2019), or take OE as an auxiliary task of AE (Wang et al., 2017; Li et al., 2018b). In order to perform ABSA for prac- tical use, the separate methods need to be pipelined together. Recently, several studies attempt to solve ABSA in a unified framework (Wang et al., 2018a; Li et al., 2019; He et al., 2019; Luo et al., 2019). Despite their effectiveness, we argue that these methods are not sufficient to yield satisfactory per- formance for the complete ABSA task. The key reason is that the interactive relations among differ- ent subtasks have been largely neglected in existing studies. These relations convey collaborative sig- nals which can enhance the subtasks in a mutual way. …

Content truncated to 500 words