close
Jump to content

Stance detection

Add links
From Wikipedia, the free encyclopedia

Stance detection is a task in natural language processing (NLP) and computational linguistics that uses content analysis and text mining to identify an author's position or stance on a specific subject. The task is most commonly framed as a classification or textual entailment problem over a fixed label set. Applications of stance detection include political discourse analysis, public health monitoring, misinformation identification, and content moderation on social media.[1]

Research interest in the task grew after the release of the SemEval-2016 Task 6 benchmark,[2][3] the first standardized dataset for stance detection on social media, which established shared evaluation norms for the NLP community.[3][4] Although historically conflated with sentiment analysis, stance detection has increasingly been recognized as a distinct task of argument mining[5][6][7] and opinion mining within NLP.[1][4][7][8]

Methodological approaches have included traditional machine learning and deep learning architectures, as well as large language models due to their ability to generalize across targets and domains.[8][9][10][11]

Methods

[edit]

Models

[edit]

Supervised classification

[edit]

Early stance detection systems relied on supervised classification using engineered features, including n-grams, sentiment lexicons, syntactic relations, and topic-relatedness measures.[12][13] Support vector machines (SVM) emerged as a dominant baseline, with SVM-based systems consistently competitive across benchmark datasets such as SemEval-2016, where topic-specific SVM models were trained on word and character n-grams.[13][14] A recurring challenge was that topic-specific features learned on one set of targets did not generalize to unseen ones, with cross-dataset evaluations showing performance drops when models were applied to data outside their training domain.[12][13][14]

Deep learning

[edit]

Deep learning approaches supplanted feature-based methods as the primary methodology in stance detection, driven by the ability to capture semantic and contextual information directly from text without manual feature engineering.[9][13] Early neural architectures applied to stance detection included convolutional neural networks for extracting local n-gram-like patterns and recurrent models such as LSTMs and bidirectional LSTMs for capturing sequential dependencies across longer contexts.[9][13]

Transformer architectures, such as BERT and its variants, marked a further shift with contextual word embeddings that improved performance across benchmark datasets such as SemEval-2016.[9][13] Unlike earlier static embedding models such as Word2Vec and GloVe, which assigned a single fixed vector to each word regardless of context, BERT uses bidirectional self-attention to produce representations that vary depending on surrounding words, enabling more precise modeling of the target-stance relationship.[15][16] BERT is typically adapted to stance detection through fine-tuning, in which a pretrained model is further trained on a labeled stance dataset, allowing it to leverage broad linguistic knowledge acquired during pretraining while specializing to the classification task.[17][18] A consistent finding across transformer-based models is that explicitly providing target information alongside the input text improves classification performance, as stance is inherently relational and cannot be reliably inferred from text alone.[15][16]

A challenge across transformer-based approaches is cross-target generalizability, as models fine-tuned on one set of targets often rely on correlations with target-specific vocabulary and fail to transfer to unseen ones, motivating approaches that inject external contextual knowledge to improve robustness.[14][16] A particular approach using encoder models is natural language inference (NLI), in which a document is paired with a hypothesis statement and the model predicts whether the document entails that hypothesis, enabling zero-shot and few-shot classification across tasks without additional training.[9][8][15]

Ensemble methods combining fine-tuned transformer models with generative large language models have also been explored as a means of leveraging the complementary strengths of discriminative and generative approaches, with voting-based ensembles demonstrating competitive performance on domain-specific stance benchmarks.[9][18] Smaller open-source encoder models trained on domain-specific NLI data can match or exceed large proprietary generative models for well-defined classification tasks while offering greater computational efficiency.[15][18][19]

Large language models

[edit]

Large language models (LLMs) have been increasingly employed for stance detection.[8][10][11] While encoder models like BERT create dense semantic representations of text that can be fine-tuned for classification, generative models produce text in response to natural-language prompts describing the task.[8][11][15]

Generative LLMs are applied to stance detection primarily through prompt engineering, with several strategies developed to improve performance. Standard zero-shot prompting asks the model to assign a label given a task description, while few-shot prompting augments the prompt with labeled examples.[11][20] Chain-of-thought prompting, which asks models to reason through intermediate steps before providing a final stance judgment.[11][20][21][22] Chain of Stance is a variation which decomposes stance detection into sequential assertions about context, viewpoint, emotion, and logical consistency before arriving at a conclusion.[10][20] Agentic and multi-expert frameworks extend this further by assigning specialized reasoning roles to different LLM agents whose outputs are then aggregated by a meta-judge.[21] Retrieval-augmented approaches similarly inject external background knowledge into the reasoning process to handle targets that require up-to-date world knowledge.[10][21][23]

Despite strong performance, generative models often cannot be archived for replication, raise concerns about political and demographic bias, and are costly to deploy at scale.[8][10][11] Performance varies substantially across datasets, target subjects, and model versions, and no single model has been found to consistently dominate across tasks.[19]

Target relationships

[edit]

The target method defines the relationship between training and evaluation targets.[10] In-target stance detection is a supervised setting in which models are trained and evaluated on the same target.[23][24] Cross-target detection relaxes this constraint, training on one set of targets and testing on unseen ones, enabling generalization to new topics without additional annotation.[16][25][26][27] Multi-target stance detection addresses the case where a single text may express positions toward multiple targets simultaneously, posing a more complex structured prediction problem.[28][29][30]

Benchmarks like the VAST (Varied Stance),[31] Stanceosaurus,[32][33] MT-CSD (Multi-Turn Conversation Stance Detection),[34][35] MmMtCSD (Multimodal Multi-turn Conversation Stance Detection),[36] and ZS-CSD (Zero-Shot Conversational Stance Detection) datasets have been used to evaluate cross-target stance detection tasks.[37]

Learning settings

[edit]

Zero-shot and few-shot learning have become central to modern stance detection as researchers seek to move beyond task-specific models requiring large labeled datasets.[38][39][40]

In zero-shot settings, a model classifies stance toward targets it has never explicitly trained on, typically by framing the task as natural language inference, or using instruction-tuned LLMs with a prompt describing the target and candidate stances.[21][31][37][38]

Few-shot learning extends this by providing a small number of labeled examples, either in-context or via lightweight fine-tuning.[8][41][42] This allows the model to adapt to new targets or domains without full retraining.[39][40][41][42]

See Also

[edit]

References

[edit]
  1. 1 2 Küçük, Dilek; Can, Fazli (2020-02-06). "Stance Detection: A Survey". ACM Comput. Surv. 53 (1): 12:1–12:37. doi:10.1145/3369026. ISSN 0360-0300.
  2. ↑ Mohammad, Saif; Kiritchenko, Svetlana; Sobhani, Parinaz; Zhu, Xiaodan; Cherry, Colin (June 2016). "SemEval-2016 Task 6: Detecting Stance in Tweets". Proceedings of the 10th International Workshop on Semantic Evaluation (SemEval-2016). SemEval. San Diego, California: Association for Computational Linguistics. pp. 31–41. doi:10.18653/v1/S16-1003. S16-1003.
  3. 1 2 ALDayel, Abeer; Magdy, Walid (2021). "Stance detection on social media: State of the art and trends". Information Processing & Management. 58 (4) 102597. doi:10.1016/j.ipm.2021.102597.
  4. 1 2 Wang, Rui; Zhou, Deyu; Jiang, Mingmin; Si, Jiasheng; Yang, Yang (2019). "A Survey on Opinion Mining: From Stance to Product Aspect". IEEE Access. 7: 41101–41124. doi:10.1109/ACCESS.2019.2906754. ISSN 2169-3536.
  5. ↑ Sobhani, Parinaz; Inkpen, Diana; Matwin, Stan (2015). "From Argumentation Mining to Stance Classification". Proceedings of the 2nd Workshop on Argumentation Mining. Association for Computational Linguistics: 67–77. doi:10.3115/v1/W15-0509.
  6. ↑ Hasan, Kazi Saidul; Ng, Vincent (2014). "Why are You Taking this Stance? Identifying and Classifying Reasons in Ideological Debates". Proceedings of the 2014 Conference on Empirical Methods in Natural Language Processing (EMNLP). Association for Computational Linguistics: 751–762. doi:10.3115/v1/D14-1083.
  7. 1 2 Boltužić, Filip; Šnajder, Jan (2014). "Back up your Stance: Recognizing Arguments in Online Discussions". Proceedings of the First Workshop on Argumentation Mining. Association for Computational Linguistics: 49–58. doi:10.3115/v1/W14-2107.
  8. 1 2 3 4 5 6 7 Burnham, Michael (2024-09-17). "Stance detection: a practical guide to classifying political beliefs in text". Political Science Research and Methods. 13 (3): 611–628. doi:10.1017/psrm.2024.35. ISSN 2049-8470.
  9. 1 2 3 4 5 6 Gera, Parush; Neal, Tempestt (2025-09-04). "Deep Learning in Stance Detection: A Survey". ACM Comput. Surv. 58 (1): 26:1–26:37. doi:10.1145/3744641. ISSN 0360-0300.
  10. 1 2 3 4 5 6 Pangtey, Lata; Bhatnagar, Anukriti; Bansal, Shubhi; Dar, Shahid Shafi; Kumar, Nagendra (2026-01-19), Large Language Models Meet Stance Detection: A Survey of Tasks, Methods, Applications, Challenges and Future Directions, arXiv, doi:10.48550/arXiv.2505.08464, arXiv:2505.08464, retrieved 2026-06-11
  11. 1 2 3 4 5 6 Griswold, Max; Robbins, Michael W.; Pollard, Michael S. (2025-12-17). "Stay Tuned: Improving Sentiment Analysis and Stance Detection Using Large Language Models". Political Analysis: 1–20. doi:10.1017/pan.2025.10023. ISSN 1047-1987.
  12. 1 2 Bar-Haim, Roy; Bhattacharya, Indrajit; Dinuzzo, Francesco; Saha, Amrita; Slonim, Noam (2017). Lapata, Mirella; Blunsom, Phil; Koller, Alexander (eds.). "Stance Classification of Context-Dependent Claims". Proceedings of the 15th Conference of the European Chapter of the Association for Computational Linguistics: Volume 1, Long Papers. Valencia, Spain: Association for Computational Linguistics: 251–261.
  13. 1 2 3 4 5 6 Cao, Rong; Luo, Xiangyang; Xi, Yaoyi; Qiao, Yaqiong (2022). "Stance detection for online public opinion awareness: An overview". International Journal of Intelligent Systems. 37 (12): 11944–11965. doi:10.1002/int.23071. ISSN 1098-111X.
  14. 1 2 3 Ng, Lynnette; Carley, Kathleen (2022-11-01). "Is my stance the same as your stance? A cross validation study of stance detection datasets". Information Processing & Management. 59 (6). doi:10.1016/j.ipm.2022.103070. ISSN 0306-4573.
  15. 1 2 3 4 5 Burnham, Michael; Kahn, Kayla; Wang, Ryan Yang; Peng, Rachel X. (2025-12-15). "Political DEBATE: Efficient Zero-Shot and Few-Shot Classifiers for Political Text". Political Analysis: 1–15. doi:10.1017/pan.2025.10028. ISSN 1047-1987.
  16. 1 2 3 4 Beck, Tilman; Waldis, Andreas; Gurevych, Iryna (2023). Palmer, Alexis; Camacho-collados, Jose (eds.). "Robust Integration of Contextual Information for Cross-Target Stance Detection". Proceedings of the 12th Joint Conference on Lexical and Computational Semantics (*SEM 2023). Toronto, Canada: Association for Computational Linguistics: 494–511. doi:10.18653/v1/2023.starsem-1.43.
  17. ↑ Karande, Hema; Walambe, Rahee; Benjamin, Victor; Kotecha, Ketan; Raghu, T. S. (2021-04-14). "Stance Detection with BERT Embeddings for Credibility Analysis of Information on Social Media". PeerJ Computer Science. 7: e467. doi:10.7717/peerj-cs.467. ISSN 2376-5992.
  18. 1 2 3 Rodriguez-Garcia, Raquel; Reyes Montesinos, Julio; Fraile-Hernandez, Jesus M.; Peñas, Anselmo (2024). Hürriyetoğlu, Ali; Tanev, Hristo; Thapa, Surendrabikram; Uludoğan, Gökçe (eds.). "HAMiSoN-Ensemble at ClimateActivism 2024: Ensemble of RoBERTa, Llama 2, and Multi-task for Stance Detection". Proceedings of the 7th Workshop on Challenges and Applications of Automated Extraction of Socio-political Events from Text (CASE 2024). St. Julians, Malta: Association for Computational Linguistics: 118–124. doi:10.18653/v1/2024.case-1.16.
  19. 1 2 Espinosa, Laura; Kebaili, Djilani; Consoli, Sergio; Kalimeri, Kyriaki; Mejova, Yelena; Salathé, Marcel (2025-03-21), Open-source solution for evaluation and benchmarking of large language models for public health, doi:10.1101/2025.03.20.25324040, retrieved 2026-06-12
  20. 1 2 3 Ma, Junxia; Wang, Changjiang; Xing, Hanwen; Zhao, Dongming; Zhang, Yazhou (2024-08-03), Chain of Stance: Stance Detection with Large Language Models, arXiv, doi:10.48550/arXiv.2408.04649, arXiv:2408.04649, retrieved 2026-06-12
  21. 1 2 3 4 Zhang, Yuanshuo; Li, Aohua; Chen, Bo; Sun, Jingbo; Zhao, Xiaobing (2026-03-14). "MSME: A Multi-Stage Multi-Expert Framework for Zero-Shot Stance Detection". Proceedings of the AAAI Conference on Artificial Intelligence. 40 (41): 34879–34887. doi:10.1609/aaai.v40i41.40791. ISSN 2374-3468.
  22. ↑ Gatto, Joseph; Sharif, Omar; Preum, Sarah M. (2023). Bouamor, Houda; Pino, Juan; Bali, Kalika (eds.). "Chain-of-Thought Embeddings for Stance Detection on Social Media". Findings of the Association for Computational Linguistics: EMNLP 2023. Singapore: Association for Computational Linguistics: 4154–4161. doi:10.18653/v1/2023.findings-emnlp.273.
  23. 1 2 Li, Ang; Liang, Bin; Zhao, Jingqian; Zhang, Bowen; Yang, Min; Xu, Ruifeng (2023). Bouamor, Houda; Pino, Juan; Bali, Kalika (eds.). "Stance Detection on Social Media with Background Knowledge". Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing. Singapore: Association for Computational Linguistics: 15703–15717. doi:10.18653/v1/2023.emnlp-main.972.
  24. ↑ Li, Yingjie; Sosea, Tiberiu; Sawant, Aditya; Nair, Ajith Jayaraman; Inkpen, Diana; Caragea, Cornelia (2021). Zong, Chengqing; Xia, Fei; Li, Wenjie; Navigli, Roberto (eds.). "P-Stance: A Large Dataset for Stance Detection in Political Domain". Findings of the Association for Computational Linguistics: ACL-IJCNLP 2021. Online: Association for Computational Linguistics: 2355–2365. doi:10.18653/v1/2021.findings-acl.208.
  25. ↑ "Cross-target stance detection: A survey of techniques, datasets, and challenges". Expert Systems with Applications. 283. 2025-07-15. doi:10.1016/j.eswa.2025.127790. ISSN 0957-4174.
  26. ↑ Ding, Daijun; Chen, Rong; Jing, Liwen; Zhang, Bowen; Huang, Xu; Dong, Li; Zhao, Xiaowen; Song, Ge (2024-01-04), Cross-target Stance Detection by Exploiting Target Analytical Perspectives, arXiv, doi:10.48550/arXiv.2401.01761, arXiv:2401.01761, retrieved 2026-06-12
  27. ↑ Xu, Chang; Paris, Cécile; Nepal, Surya; Sparks, Ross (2018). "Cross-Target Stance Classification with Self-Attention Networks". Proceedings of the 56th Annual Meeting of the Association for Computational Linguistics. Association for Computational Linguistics: 778–783. doi:10.18653/v1/P18-2123.
  28. ↑ Chen, Chenguang; Xi, Wen; Zhou, Bin (2021-01-11). "Multi-Target Stance Detection with Multi-Task Learning". Proceedings of the 2020 9th International Conference on Computing and Pattern Recognition. ICCPR '20. New York, NY, USA: Association for Computing Machinery: 111–116. doi:10.1145/3436369.3436473. ISBN 978-1-4503-8783-5.
  29. ↑ Sobhani, Parinaz; Inkpen, Diana; Zhu, Xiaodan (2017). Lapata, Mirella; Blunsom, Phil; Koller, Alexander (eds.). "A Dataset for Multi-Target Stance Detection". Proceedings of the 15th Conference of the European Chapter of the Association for Computational Linguistics: Volume 2, Short Papers. Valencia, Spain: Association for Computational Linguistics: 551–557.
  30. ↑ Wei, Penghui; Lin, Junjie; Mao, Wenji (2018-06-27). "Multi-Target Stance Detection via a Dynamic Memory-Augmented Network". The 41st International ACM SIGIR Conference on Research & Development in Information Retrieval. SIGIR '18. New York, NY, USA: Association for Computing Machinery: 1229–1232. doi:10.1145/3209978.3210145. ISBN 978-1-4503-5657-2.
  31. 1 2 Allaway, Emily; McKeown, Kathleen (2020). "Zero-Shot Stance Detection: A Dataset and Model using Generalized Topic Representations". Proceedings of the 2020 Conference on Empirical Methods in Natural Language Processing. Association for Computational Linguistics: 8913–8931. doi:10.18653/v1/2020.emnlp-main.717.
  32. ↑ Zheng, Jonathan; Baheti, Ashutosh; Naous, Tarek; Xu, Wei; Ritter, Alan (2022). "Stanceosaurus: Classifying Stance Towards Multicultural Misinformation". Proceedings of the 2022 Conference on Empirical Methods in Natural Language Processing. Association for Computational Linguistics: 2132–2151. doi:10.18653/v1/2022.emnlp-main.138.
  33. ↑ Lavrouk, Anton; Ligon, Ian; Zheng, Jonathan; Naous, Tarek; Xu, Wei; Ritter, Alan (2024). van der Goot, Rob; Bak, JinYeong; Müller-Eberstein, Max; Xu, Wei; Ritter, Alan; Baldwin, Tim (eds.). "Stanceosaurus 2.0 - Classifying Stance Towards Russian and Spanish Misinformation". Proceedings of the Ninth Workshop on Noisy and User-generated Text (W-NUT 2024). San \.Giljan, Malta: Association for Computational Linguistics: 31–43. doi:10.18653/v1/2024.wnut-1.4.
  34. ↑ Niu, Fuqiang; Yang, Min; Li, Ang; Zhang, Baoquan; Peng, Xiaojiang; Zhang, Bowen (2024). Calzolari, Nicoletta; Kan, Min-Yen; Hoste, Veronique; Lenci, Alessandro; Sakti, Sakriani; Xue, Nianwen (eds.). "A Challenge Dataset and Effective Models for Conversational Stance Detection". Proceedings of the 2024 Joint International Conference on Computational Linguistics, Language Resources and Evaluation (LREC-COLING 2024). Torino, Italia: ELRA and ICCL: 122–132.
  35. ↑ Niu, Fuqiang; Yang, Yi; Fu, Xianghua; Dai, Genan; Zhang, Bowen (2025-05-08). "C-MTCSD: A Chinese Multi-Turn Conversational Stance Detection Dataset". Companion Proceedings of the ACM on Web Conference 2025. New York, NY, USA: ACM: 769–772. doi:10.1145/3701716.3715307.
  36. ↑ Niu, Fuqiang; Cheng, Zebang; Fu, Xianghua; Peng, Xiaojiang; Dai, Genan; Chen, Yin; Huang, Hu; Zhang, Bowen (2024-10-28). "Multimodal Multi-turn Conversation Stance Detection: A Challenge Dataset and Effective Model". Proceedings of the 32nd ACM International Conference on Multimedia. MM '24. New York, NY, USA: Association for Computing Machinery: 3867–3876. doi:10.1145/3664647.3681416. ISBN 979-8-4007-0686-8.
  37. 1 2 Ding, Yuzhe; He, Kang; Li, Bobo; Zheng, Li; He, Haijun; Li, Fei; Teng, Chong; Ji, Donghong (2025). "Zero-Shot Conversational Stance Detection: Dataset and Approaches". Findings of the Association for Computational Linguistics. Association for Computational Linguistics: 3221–3235. doi:10.18653/v1/2025.findings-acl.168.
  38. 1 2 Wen, Haoyang; Hauptmann, Alexander (2023). "Zero-Shot and Few-Shot Stance Detection on Varied Topics via Conditional Generation". Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics. Association for Computational Linguistics: 1491–1499. doi:10.18653/v1/2023.acl-short.127.
  39. 1 2 Liu, Rui; Lin, Zheng; Tan, Yutong; Wang, Weiping (2021). "Enhancing Zero-shot and Few-shot Stance Detection with Commonsense Knowledge Graph". Findings of the Association for Computational Linguistics: ACL-IJCNLP 2021. Association for Computational Linguistics: 3152–3157. doi:10.18653/v1/2021.findings-acl.278.
  40. 1 2 Zhang, Jiarui; Wu, Shaojuan; Zhang, Xiaowang; Feng, Zhiyong (2023-04-30). "Task-Specific Data Augmentation for Zero-shot and Few-shot Stance Detection". Companion Proceedings of the ACM Web Conference 2023. WWW '23 Companion. New York, NY, USA: Association for Computing Machinery: 160–163. doi:10.1145/3543873.3587337. ISBN 978-1-4503-9419-2.
  41. 1 2 Jiang, Yan; Gao, Jinhua; Shen, Huawei; Cheng, Xueqi (2022-07-07). "Few-Shot Stance Detection via Target-Aware Prompt Distillation". Proceedings of the 45th International ACM SIGIR Conference on Research and Development in Information Retrieval. SIGIR '22. New York, NY, USA: Association for Computing Machinery: 837–847. doi:10.1145/3477495.3531979. ISBN 978-1-4503-8732-3.
  42. 1 2 Hardalov, Momchil; Arora, Arnav; Nakov, Preslav; Augenstein, Isabelle (2022-06-28). "Few-Shot Cross-Lingual Stance Detection with Sentiment-Based Pre-training". Proceedings of the AAAI Conference on Artificial Intelligence. 36 (10): 10729–10737. doi:10.1609/aaai.v36i10.21318. ISSN 2374-3468.
Retrieved from "https://en.wikipedia.org/w/index.php?title=Stance_detection&oldid=1362312536"