Siddha Medicine Dataset: Mapping Diseases to Traditional\u00a0Remedies

The Siddha Medicine Dataset: Mapping Diseases to Traditional Remedies is a curated collection of traditional Siddha medical knowledge containing 110 records. Each record includes disease names in both Tamil and English, corresponding medicine details, ingredients, reference book names, and original Tamil poetic lines describing the remedy. The dataset also provides medicine-type classifications in both Tamil and English. All data entries are primarily in the Tamil language, a low-resource linguistic domain, ensuring authenticity and preservation of traditional terminology. Data was manually curated with the guidance of a Siddha domain expert (second-profession student) to ensure accuracy and contextual reliability. The dataset was compiled from classical Siddha sources such as Gunapaadam: Thaadhu Vaguppu, Gunapaadam: Seeva Vaguppu, and Gunapaadam: Mooligai Vaguppu. The TV Sambasivam Pillai Dictionary was referenced to map Tamil disease names to their equivalent English terms. Machine learning models, including Random Forest and CNN, were applied to classify medicine types, achieving accuracies of 91% and 88% respectively. This dataset represents an initial effort to build high-quality, low-resource, native-language datasets for traditional medicine and aims to encourage further regional-language data collection for AI and healthcare research.