Large language model-assisted natural language processing reveals stigmatizing discourse regarding psoriasis on social media

This dataset provides supplementary materials supporting the Research Letter accepted for reconsideration in JAAD International. It contains detailed methodological, computational, and validation information omitted from the main text due to space constraints. Specifically, the dataset includes:
1.Detailed Methodology – full description of data preprocessing, traditional-to-simplified Chinese text normalization, rule-based classification schema (26 rules), and model onfiguration parameters for DeepSeek-V3.
2.Model Prompts and Validation Workflow – structured prompts, fault-tolerant parsing logic, and manual double-coding procedures used to benchmark model performance.
3.Evaluation Outputs – performance metrics (precision, sensitivity, F1-score) for stigmatization, destigmatization, and neutral categories, including confusion matrix data.