Exploiting Sentiment and Common Sense for Zero-shot Stance Detection

Proceedings of the 29th International Conference on Computational Linguistics, pages 7112–7123 October 12–17, 2022. 7112 Exploiting Sentiment and Common Sense for Zero-shot Stance Detection Yun Luo 1,2, Zihan Liu 2, Yuefeng Shi 2, Stan Z. Li 2, Yue Zhang 2,3 1 School of Computer Science And Technology, Zhejiang University, Hangzhou, 310024, P.R. China. 2 School of Engineering, Westlake University, Hangzhou, 310024, P.R. China. 3 Institute of Advanced Technology, Westlake Institute for Advanced Study, Hangzhou, 310024, P.R. China. {luoyun, guofang, liuzihan, stan.zq.li, zhangyue}@westlake.edu.cn Abstract The stance detection task aims to classify the stance toward given documents and topics. Since the topics can be implicit in documents and unseen in training data for zero-shot set- tings, we propose to boost the transferability of the stance detection model by using senti- ment and commonsense knowledge, which are seldom considered in previous studies. Our model includes a graph autoencoder module to obtain commonsense knowledge and a stance detection module with sentiment and common- sense. Experimental results show that our model outperforms the state-of-the-art meth- ods on the zero-shot and few-shot benchmark dataset–VAST. Meanwhile, ablation studies prove the significance of each module in our model. Analysis of the relations between senti- ment, common sense, and stance indicates the effectiveness of sentiment and common sense. 1 Introduction Stance detection aims to identify the authors’ at- titudes or positions (Pro (support), Con (oppose), Neu (neutral)) towards a specific target such as an entity, a topic. (Mohammad et al., 2017, 2016; Walker et al., 2012; Qiu et al., 2015; Zhang et al., 2017). It is crucial for understanding opinions and analyzing how opinions are presented in texts re- garding specific issues, and much work has been done building stance detection models (Wei et al., 2016; Dias and Becker, 2016; Allaway and Mck- eown, 2020). There are two salient challenges to the task. First, obtaining rich annotated data in stance detection is time-consuming and labor- intensive. To address this issue, Allaway and Mck- eown (2020) propose the dataset VAST containing various topics for few-shot and zero-shot stance detection tasks, requiring the model to classify the stance of topics unseen in the training set. Sec- ond, the topic is often not explicitly mentioned in the document, resulting in difficulty. Considering Text : I totally agree with this premise. As a younger per- son I was against Nuclear power (I was in college during 3 mile island) but now it seems that nuclear should be in the mix. Fission technology is better, and will continue to get better if we actively promote its development. The pro- spect of fusion energy also needs to be explored. If it's good enough for the sun and the stars, it's good enough for me. Text : Example 2 Topic : Nuclear power Stance : Pro Example 1 Topic : Olympics Stance : Pro The games should proceed. Athletes have made tremendous sacrifices to qualify and be prepared. It would be cruel to deny them their chance. In the future the Games should be held in …

Content truncated to 500 words