SRQ: Stance in Twitter Replies and Quotes on Controversial Issues

Stance in Replies and Quotes (SRQ) is a human-labeled dataset for stance detection in Twitter conversations, created by Ramón Villa-Cox at Carnegie Mellon University. It contains over 5200 stance labels from responses to posts on controversial issues, focusing on both replies and quotes. The dataset's collection methodology was designed to favor the selection of denial-type responses.

Use Cases

• Training stance detection models based on labeled replies and quotes from Twitter.
• Identifying rumors and antagonistic user relationships based on the dataset's focus on denial-type responses.
• Evaluating model generalization across different social media conversation types, as mentioned in the description.
• Advancing opinion mining research on controversial topics using a large, human-labeled corpus.

Strengths

• Over 5200 human-labeled stance annotations, which the description states is the largest such dataset for Twitter conversations.
• Explicitly distinguishes between reply and quote conversation types on Twitter.
• Collection methodology designed to favor denial-type responses, a class noted as useful for rumor identification.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count and file format are unknown, which may limit suitability assessment.
• Last update date is unknown; freshness unverified.

Provenance

Source

Carnegie Mellon University

Collection Method

Human labeling of stance in responses to Twitter posts on controversial issues.