Sarcastic Tweets with GIF Reactions for Affect Prediction

Featuring 30,000 sarcastic tweets paired with GIF reactions. It was created for research on predicting induced affect, as detailed in an ACL 2021 paper by Shmueli, Ray, and Ku.

Use Cases

• Train a text classification model to predict the appropriate GIF reaction for a given sarcastic tweet.
• Analyze the relationship between sarcastic tweet content and the affective labels of the paired GIFs.
• Build a model to detect sarcasm in tweets using the GIF reaction as an indirect supervision signal.
• Study patterns in crowdsourced annotations linking textual sarcasm to visual-emotional GIF responses.

Strengths

• Contains 30,000 data points, providing a substantial sample for analysis.
• Specifically focuses on sarcastic tweets, a distinct and challenging linguistic phenomenon.
• Pairs text with GIF reactions, enabling multimodal analysis of affect.
• Associated with a peer-reviewed ACL 2021 research paper.

Limitations

• The dataset size of 30K rows may be limited for training large, data-hungry neural models.
• Annotations are crowdsourced, which can introduce label noise or subjective bias.
• Geographic coverage is indicated as US region, limiting generalizability to other cultures and languages.

Provenance

Source

Original repository at https://github.com/bshmueli/ReactionGIF

Collection Method

Crowdsourced annotations linking sarcastic tweets to GIF reactions.

Freshness

Last updated on the platform in September 2022.

Geography

Primarily US region, as indicated by tags.