HSQ-TD:Health & Spiritual Qigong Knowledge Dataset for Large Language Model Fine-Tuning

Read full text Summarise with VIRAT
Authors: Dong Zhaozhao, Li Ailin
Cited by: 0

Summary

No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.

Summarise with VIRAT Free to try with your Science Catalogs account.

Abstract

Comprising 15,604 data instances stored in JSONL format, the HSQ-TD dataset is a high-quality Named Entity Recognition (NER) dataset specifically constructed for fine-tuning Large Language Models (LLMs) on corpora encompassing fitness and wellness news as well as ancient Qigong texts. Each data instance consists of three attributes: instruction, input, and output. The instruction designates the model as a text analysis expert in the domain of fitness and wellness Qigong, explicitly delineating the definitions and strict output templates for eight core entity categories (Gongfa, Action, Body Part, Acupoint, Meridian, Substance, Effect, and Symptom) to compel the model to focus on specific entity extraction rules and task boundaries. The input serves as the target object for the model's information extraction and analytical reasoning, consisting of the original corpus text rigorously sanitized via regular expressions and rule engines to eliminate modern medical vocabulary contamination and overlapping boundary conflicts. Finally, the output represents the structured response generated by the model's reasoning based on the instruction and input text, strictly returning the extracted entities in a pure JSON format.

Fields of Study & Subjects

Topic Modeling Biomedical Text Mining and Ontologies Traditional Chinese Medicine Studies

Publication details

Journal / source
ScienceDB
Publication date
14 March 2026
Language
ZH
Document type
Dataset