Transferable End-to-End Aspect-based Sentiment Analysis with Selective Adversarial Learning

Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9th International Joint Conference on Natural Language Processing, pages 4590–4600, Hong Kong, China, November 3–7, 2019. c⃝2019 Association for Computational Linguistics 4590 Transferable End-to-End Aspect-based Sentiment Analysis with Selective Adversarial Learning Zheng Li1, Xin Li2, Ying Wei3, Lidong Bing4, Yu Zhang5, Qiang Yang1 1The Hong Kong University of Science and Technology, Hong Kong 2The Chinese University of Hong Kong, Hong Kong 3Tencent AI Lab, Shenzhen, China 4R&D Center Singapore, Machine Intelligence Technology, Alibaba DAMO Academy 5Southern University of Science and Technology, Shenzhen, China {zlict, qyang}@cse.ust.hk, lixin@se.cuhk.edu.hk, judywei@tencent.com l.bing@alibaba-inc.com, yu.zhang.ust@gmail.com Abstract Joint extraction of aspects and sentiments can be effectively formulated as a sequence label- ing problem. However, such formulation hin- ders the effectiveness of supervised methods due to the lack of annotated sequence data in many domains. To address this issue, we firstly explore an unsupervised domain adap- tation setting for this task. Prior work can only use common syntactic relations between aspect and opinion words to bridge the do- main gaps, which highly relies on external linguistic resources. To resolve it, we pro- pose a novel Selective Adversarial Learning (SAL) method to align the inferred correla- tion vectors that automatically capture their latent relations. The SAL method can dy- namically learn an alignment weight for each word such that more important words can pos- sess higher alignment weights to achieve fine- grained (word-level) adaptation. Empirically, extensive experiments1 demonstrate the effec- tiveness of the proposed SAL method. 1 Introduction End-to-End Aspect-Based Sentiment Analysis (E2E-ABSA) aims to jointly detect the aspect terms explicitly mentioned in sentences and pre- dict the sentiment polarities over them (Liu, 2012; Pontiki et al., 2014). For example, in the sentence “The AMD Turin Processor seems to always per- form much better than Intel”, the user mentions two aspect terms, i.e., “AMD Turin Processor” and “Intel”, and expresses positive and negative sentiments over them, respectively. Typically, prior work formulates E2E-ABSA as a sequence labeling problem over a unified tagging scheme (Mitchell et al., 2013; Zhang et al., 2015; Li et al., 2019a). The unified tagging scheme 1The code is available at https://github.com/ hsqmlzno1/Transferable-E2E-ABSA connects a set of aspect boundary tags (e.g., {B, I, E, S, O} denotes the beginning of, inside of, end of, single-word, and no aspect term), and sentiment tags (e.g. {POS, NEG, NEU} denotes positive, negative or neutral sentiment) together to constitute a joint label space for each word. As such, “AMD Turin Processor” and “Intel” should be tagged with {B-POS, I-POS, E-POS} and {S-NEG}, respectively, while the remaining words are tagged with O. This formulation makes two sub-tasks joint modeling easier, and mean- while, tend to be low-resource. There usually exist few annotated data for each new domain, where la- beling each word with a unified tag could be more time-consuming and expensive. To alleviate the dependence on domain supervi- sions, we explore an unsupervised domain adapta- tion setting for E2E-ABSA, which aims to lever- age knowledge from a …

Content truncated to 500 words