Year: 2022
Type: conference-paper
Abstract: We present Stanceosaurus, a new corpus of 28,033 tweets in English, Hindi, and Arabic annotated with stance towards 251 misinformation claims.As far as we are aware, it is the largest corpus annotated with stance towards misinformation claims.The claims in Stanceosaurus originate from 15 fact-checking sources that cover diverse geographical regions and cultures.Unlike existing stance datasets, we introduce a more fine-grained 5class labeling strategy with additional subcategories to distinguish implicit stance.Pretrained transformer-based stance classifiers that are fine-tuned on our corpus show good generalization on unseen claims and regional claims from countries outside the training data.Cross-lingual experiments demonstrate Stanceosaurus' capability of training multilingual models, achieving 53.1 F1 on Hindi and 50.4 F1 on Arabic without any targetlanguage fine-tuning.Finally, we show how a domain adaptation method can be used to improve performance on Stanceosaurus using additional RumourEval-2019 data.We make Stanceosaurus publicly available to the research community and hope it will encourage further work on misinformation identification across languages and cultures.1 Source Country & Regions Lang #Claims #Tweets Irr.Sup.Ref. Dis.Que.Snopes USA (80%), INT'L (16.7%),Other (3.3%) en 30 3197 1051 428 229 1447 42 Poynter Europe (5%), INT'L (90%), Other (5%) en 20 2197 949 274 97 844 33 FullFact UK (30%), INT'L (55%), Other (15%) en 20 2379 806 300 179 1057 37 AFP Fact Check CAN Canada (55%), INT'L (30%), Other (15%) en 20 2078 746 252 130 910 40 AAP Fact Check Australia (10%), INT'L (65%), Other (25%) en 20 2302 739 374 136 1019 34 AFP Fact Check NZ New Zealand (15%), INT'L (75%), Other (10%) en 20 2227 879 194 81 1044 29 Blackdotresearch Singapore (30%), INT'L (55%), Other (15%) en 20 2307 842 248 113 1076 28 Factly India (45%), INT'L (55%) en 20 1979 889 190 117 734 49 Politifact USA (20%), INT'L (35%), Other (45%) en 20 2041 984 289 8 753 7 Alt News India (90.4%),INT'L (4.8%),Other (4.8%) hi 21 1730 550 489 172 500 19 Aajtak India (67%), Other (33%) hi 9 806 456 110 40 193 7 Hindi Newschecker India (56%), Other (44%) hi 9 781 195 313 46 219 8 MISBAR Arab World (58.3%),INT'L (8.3%),Other (33.4%) ar 12 2283 454 514 203 1031 81 Fatabyyano Arab World (28.5%),INT'L (57.1%),Other (14.4%) ar 7 986 234 163 49 522 18 Maharat Fact-o-meter INT'L (100%) ar 3 740
Institution: Georgia Institute of Technology
Language: English
Subfield: Sociology and Political Science
Field: Social Sciences
Domain: Social Sciences
Open Access status: hybrid
OpenAlex ID: W4385572920