Manually Annotated COVID-19 Vaccine Attitude Tweets

Aggregating 2,564 manually annotated English tweets related to COVID-19 vaccines, curated to maximize the likelihood of strong emotional tone. It was created by webimmunization for text classification tasks, with tweets labeled into three attitude classes.

Use Cases

• Train a text classifier to predict the PRO, ANTI, or NEUTRAL attitude label from tweet text.
• Analyze the distribution of manually annotated attitude labels across the 2,564 tweets.
• Use the curated tweet text for sentiment analysis focused on COVID-19 vaccine discourse.

Strengths

• Dataset consists of 2,564 manually annotated data points.
• Tweets were curated to maximize the likelihood of containing a strong emotional tone.
• Annotations are crowdsourced, providing multiple perspectives on labeling.

Limitations

• Dataset size of 2,564 rows is relatively small for training complex deep learning models.
• Geographic coverage may be biased, as tags indicate a 'us' region focus.
• Data is from a specific period (last updated 2022) and may not reflect current vaccine attitudes.

Provenance

Source

webimmunization on Hugging Face.

Collection Method

Manually annotated tweets, crowdsourced.

Freshness

Last updated on 2022-10-25.

Geography

Primarily US region, as indicated by tags.