Semantics-Preserved Data Augmentation for Aspect-Based Sentiment Analysis

Proceedings of the 2021 Conference on Empirical Methods in Natural Language Processing, pages 4417–4422 November 7–11, 2021. c⃝2021 Association for Computational Linguistics 4417 Semantics-Preserved Data Augmentation for Aspect-Based Sentiment Analysis Ting-Wei Hsu 1, Chung-Chi Chen 1, Hen-Hsen Huang 2,3, Hsin-Hsi Chen 1,3 1 Department of Computer Science and Information Engineering, National Taiwan University, Taiwan 2 Institute of Information Science, Academia Sinica, Taiwan 3 MOST Joint Research Center for AI Technology and All Vista Healthcare, Taiwan {twhsu,cjchen}@nlg.csie.ntu.edu.tw, hhhuang@iis.sinica.edu.tw, hhchen@ntu.edu.tw Abstract Both the issues of data deficiencies and seman- tic consistency are important for data augmen- tation. Most of previous methods address the first issue, but ignore the second one. In the cases of aspect-based sentiment analysis, vi- olation of the above issues may change the aspect and sentiment polarity. In this paper, we propose a semantics-preservation data aug- mentation approach by considering the impor- tance of each word in a textual sequence ac- cording to the related aspects and sentiments. We then substitute the unimportant tokens with two replacement strategies without altering the aspect-level polarity. Our approach is evaluated on several publicly available senti- ment analysis datasets and the real-world stock price/risk movement prediction scenarios. Ex- perimental results show that our methodology achieves better performances in all datasets. 1 Introduction Data annotation, the first step of most artificial in- telligence researches, often takes lots of time and is very expensive. Many studies propose method- ologies for augmenting data based on a few anno- tations (Wei and Zou, 2019; Jiao et al., 2020; Xie et al., 2020; Wu et al., 2019; Kobayashi, 2018) to reduce the cost of annotation. However, most of them encounter a problem—hardly ensuring read- ability and semantic coherence. To overcome these problems, we introduce a novel method, selective perturbed masking (SPM), to measure the impor- tance of each word in a textual sentence, and further replace the unimportant word with pre-trained lan- guage models. In this way, the semantics of a given instance will be preserved, and the training set will be enlarged with various auto-generated data. While testing the data augmentation methods, classification tasks are often adopted. Sentiment analysis is one of the well-known classification tasks. To mining more fine-grained information, aspect-based sentiment analysis (ABSA) is pro- posed. ABSA not only aims at detecting the senti- ment polarity but also attempts to mine the analysis aspect. It can be extended to subtasks of aspect category sentiment classification (ACSC), aspect term sentiment classification (ATSC), and aspect term extraction (ATE). For example, the answers to these tasks to the given sentence “the staff was so kind” are (service, positive), (staff, positive), and “staff”, respectively. These tasks are commonly taken as examples for evaluating the performance of augmentation methods. In this paper, we explore both sentiment analysis and all subtasks in ABSA to show the usefulness of the proposed method in semantics preservation. In addition to probe on sentiment analysis and ABSA tasks, we also experiment on stock price and risk movement prediction tasks. These experiments have always been considered …

Content truncated to 500 words