{"title":"An Heuristics-Based, Weakly-Supervised Approach for Classification of Stance in Tweets","authors":"Marcelo Dias, Karin Becker","doi":"10.1109/WI.2016.0021","DOIUrl":null,"url":null,"abstract":"Stance detection is the task of automatically identifying if the text author is in favor or against a subject or target. This paper presents a weakly supervised approach for stance detection in tweets based solely on their contents. The approach relies on a set of heuristics used to automatically label tweets with regard to stance, which has a twofold purpose: a) automatic creation of a training corpus to develop a predictive model using a supervised learning algorithm, and b) to complement the predictive model when determining the stance of tweets. The paper analyzes the performance of the approach considering six distinct stance targets. We achieved promising results, with weighted F-measure varying from 52% to 67%.","PeriodicalId":6513,"journal":{"name":"2016 IEEE/WIC/ACM International Conference on Web Intelligence (WI)","volume":"3 1","pages":"73-80"},"PeriodicalIF":0.0000,"publicationDate":"2016-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"2","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"2016 IEEE/WIC/ACM International Conference on Web Intelligence (WI)","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/10.1109/WI.2016.0021","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"","JCRName":"","Score":null,"Total":0}
引用次数: 2
Abstract
Stance detection is the task of automatically identifying if the text author is in favor or against a subject or target. This paper presents a weakly supervised approach for stance detection in tweets based solely on their contents. The approach relies on a set of heuristics used to automatically label tweets with regard to stance, which has a twofold purpose: a) automatic creation of a training corpus to develop a predictive model using a supervised learning algorithm, and b) to complement the predictive model when determining the stance of tweets. The paper analyzes the performance of the approach considering six distinct stance targets. We achieved promising results, with weighted F-measure varying from 52% to 67%.