Chen C.-CTsai H.-YHuang H.-HHSIN-HSI CHEN2023-06-092023-06-092020https://www.scopus.com/inward/record.uri?eid=2-s2.0-85114437676&doi=10.1109%2fWIIAT50758.2020.00047&partnerID=40&md5=7435219ef1c4f6d45c18e81337da5153https://scholars.lib.ntu.edu.tw/handle/123456789/632523Stance detection has attracted attention for several years. Previous work focuses mainly on a supervised topic-specific setting which requires labeled data for each individual topic. In this paper, we discuss the characteristics of different types of topics, and the interaction among sentiment, target, and stance in a sentence. We propose an approach without the need of stancelabeled data to identify stance incorporating the findings of their interaction. The proposed approach is topic independent and can be applied to individual topics flexibly. Furthermore, we evaluate our method on the SemEva1-2016 dataset for detecting stance in tweets, which contains six topics of two different types. Experimental results show that our approach is promising even when stance-labeled data is not available. © 2020 IEEE.Capsule network; Sentiment analysis; Stance detection[SDGs]SDG16Intelligent agents; Labeled dataconference paper10.1109/WIIAT50758.2020.000472-s2.0-85114437676