Enhancing Zero-shot and Few-shot Stance Detection with Commonsense Knowledge Graph

Findings of the Association for Computational Linguistics: ACL-IJCNLP 2021, pages 3152–3157 August 1–6, 2021. ©2021 Association for Computational Linguistics 3152 Enhancing Zero-shot and Few-shot Stance Detection with Commonsense Knowledge Graph Rui Liu1,2, Zheng Lin1∗, Yutong Tan1,2, Weiping Wang1 1Institute of Information Engineering, Chinese Academy of Sciences 2School of Cyber Security, University of Chinese Academy of Sciences {liurui1995,linzheng,tanyutong,wangweiping}@iie.ac.cn Abstract In this paper, we consider a realistic scenario on stance detection with more application po- tential, i.e., zero-shot and few-shot stance de- tection, which identifies stances for a wide range of topics with no or very few training ex- amples. Conventional data-driven approaches are not applicable to the above zero-shot and few-shot scenarios. For human beings, com- monsense knowledge is a crucial element of understanding and reasoning. In the absence of annotated data and cryptic expression of users’ stance, we believe that introducing com- monsense relational knowledge as support for reasoning can further improve the generaliza- tion and reasoning ability of the model in the zero-shot and few-shot scenarios. Specifically, we introduce a commonsense knowledge en- hanced model to exploit both the structural- level and semantic-level information of the re- lational knowledge. Extensive experiments demonstrate that our model outperforms the state-of-the-art methods on zero-shot and few- shot stance detection task. 1 Introduction Stance detection aims to identify the text authors’ attitudes or positions towards a specific topic as a category label from this set: {Pro, Con, Neu- tral} (Mohammad et al., 2016b, 2017). Conven- tionally, this task is designed to learn a target- specific classifier for prediction on the same topic. Afterward, cross-target stance detection comes out as a subclass of the initial generic stance detec- tion, where the classifier is adapted from differ- ent but closely related topics (e.g., training classi- fier on ”Hillary Clinton” and predicting on ”Don- ald Trump”) (Augenstein et al., 2016a). However, both target-specific and cross-target stance detec- tion models (Du et al., 2017; Wei and Mao, 2019; ∗Zheng Lin is the corresponding author. Topic: Stability Stance: Pro Text: Tenure does not mean a teacher cannot lose their job. It requires due process before termination. Before tenure is achieved, a teacher can be fired without due process. In the Atlanta School District administrators, fearing that low test scores would cost them their jobs, instructed teachers to change student test responses. Without tenure and due process, teachers risked being fired if they didn’t follow instructions. tenure score job change termination stability low test cost fire RelatedTo RelatedTo IsA stable Antonym RelatedTo RelatedTo PartOf perturbation RelatedTo Antonym RelatedTo HasSubevent IsA RelatedTo teacher RelatedTo Figure 1: An example where the topic isn’t contained in the text. Entity mentions in the text and the topic are highlighted. We omit the reverse edges in the relation graph for clarity. Augenstein et al., 2016a; Zhang et al., 2020) require a large number of training examples with manual annotation, and annotating data for thousands of new topics is time-consuming and expensive. In this paper, we focus on zero-shot and few-shot stance detection (Allaway and …

Content truncated to 500 words