Chinese COVID-19 Weibo Posts With Six Emotion Labels

8,606 Weibo posts related to the COVID-19 epidemic are labeled into six emotion categories: neutral, happy, angry, sad, fear, and surprise. The dataset includes a validation set of 2,000 posts and a test set. It was created by author 'souljoy' and last updated in December 2022.

Use Cases

• Train a text classifier to predict the six emotion labels (neutral, happy, angry, sad, fear, surprise) from Weibo post content.
• Analyze the distribution of specific emotions like 'fear' or 'anger' within COVID-19 related social media discourse.
• Fine-tune a sentiment analysis model on Chinese social media text using the 'happy' and 'sad' labels as supervision signals.
• Study temporal shifts in public emotion (e.g., 'surprise') by linking posts to the epidemic timeline if metadata is available.

Strengths

• Contains 8,606 labeled training samples for model development.
• Emotion labels cover six distinct psychological categories for detailed analysis.
• Includes separate validation (2,000 posts) and test sets for proper evaluation.

Limitations

• Dataset size is modest for deep learning, with under 10,000 total training posts.
• Emotion labels may be subjective and could contain annotation noise.
• Limited to Chinese-language posts from a single platform (Weibo), restricting geographic and linguistic generalizability.

Provenance

Source

Weibo (Chinese microblogging platform).

Collection Method

Posts were obtained by filtering epidemic-related content using relevant keywords during the COVID-19 pandemic.

Freshness

Last updated on 2022-12-29; data collection period is not specified but likely covers the active COVID-19 epidemic phase.

Geography

Presumably China, as Weibo is a primarily Chinese platform.