Stance Detection

Stance detection is the task of determining the position or stance that a statement (typically a social media post or reply) takes with respect to a claim or proposition. Rather than evaluating truth, stance classification identifies how a speaker orients toward a claim—whether they support it, deny it, question it, or simply comment on it.

Stance categories

The widely-adopted SDQC taxonomy (Support/Deny/Query/Comment) comprises:

• Support: The statement endorses, agrees with, or affirms the claim
• Deny: The statement contradicts, refutes, or disagrees with the claim
• Query: The statement questions, seeks clarification about, or expresses doubt toward the claim
• Comment: The statement is relevant to the claim but does not take a clear stance toward its truth

Applications

• Rumour verification: Understanding community response to unverified claims (stance often predicts veracity)
• Argument mining: Extracting pro and con arguments on controversial topics
• Propaganda detection: Identifying persuasive framing of politically-charged claims
• Debate analysis: Mapping argumentative positions in online discussions

Key challenges

• Context dependency: Classifying stance requires understanding the claim and conversational context
• Implicit stance: Replies may express stance indirectly through sarcasm, negation, or figurative language
• Class imbalance: Comment/irrelevant posts dominate social media, making support/deny/query relatively rare
• Cross-platform variation: Discourse norms and response patterns differ between Twitter, Reddit, news comment sections, etc.

Key papers

• Hanselowski et al. (2019) — A Richly Annotated Corpus for Different Tasks in Automated Fact-Checking — stance detection on 16,509 evidence text snippets from heterogeneous web sources with three-class labels (support/refute/no stance); Cohen's Kappa κ = 0.7 inter-annotator agreement; best-performing model (AtheneMLP) reaches F1 = 0.596
• Zubiaga et al. (2016) — Stance Classification in Rumours as a Sequential Task Exploiting the Tree Structure of Social Media Conversations — demonstrates that exploiting conversational tree structure improves stance classification; introduces Tree CRF for modeling rumour conversations
• A Survey on Stance Detection for Mis- and Disinformation Identification — comprehensive survey of stance detection for misinformation
• RumourEval 2019: Determining Rumour Veracity and Support for Rumours — RumourEval shared task on stance and veracity
• SemEval-2017 Task 8: RumourEval: Determining rumour veracity and support for rumours — Original RumourEval shared task design
• The Fact Extraction and VERification (FEVER) Shared Task — FEVER task combining evidence retrieval and stance

See also

• Rumour Verification — stance classification as a component of the rumour resolution pipeline
• Propagation-based fake news detection — using stance distribution in conversation threads