Cancer Peer-Support Community Posts with Emotional Tone Labels

A dataset of patient-authored and caregiver-authored text from online cancer peer-support communities, annotated for emotional tone. The dataset includes original labels and parallel AI-generated labels from a large language model, along with structured context variables. It was created by Shuo Xu and last updated on May 29, 2026.

Use Cases

• Training supervised models for emotional tone classification based on the three-class target (Negative/Neutral/Positive).
• Comparing human and AI annotation quality based on the parallel LLM-generated labels.
• Evaluating the impact of context feature augmentation on model performance using the prepended reporter role and cancer type variables.
• Analyzing systematic measurement shifts introduced by LLM-based supervision in label prevalence.

Strengths

• Dataset is openly licensed under CC-BY-4.0.
• Includes both original human labels and parallel AI-generated labels for comparison.
• Explicit context features (reporter role and cancer type) are extracted and provided.

Limitations

• The dataset is very small at 5.5 KB, indicating a limited scope.
• Row count and column-level documentation are unknown, requiring manual inspection after download.
• The description focuses on a modeling study; the exact content and structure of the data file require verification.

Provenance

Source

Derived from the 'Mental Health Insights: Vulnerable Cancer Survivors & Caregivers' dataset.

Collection Method

Text was sourced from online cancer peer-support communities. Labels were derived from four original categories collapsed to three, with additional AI labels generated by an LLM.

Freshness

Last updated 2026-05-29 17:34:07.