CMQA20K: Dataset Construction for Chinese Medicine QA Systems Using Artificial Intelligence

Artificial intelligence is driving the inheritance and innovation across various industries, with automatic question generation technology playing a key role in the exploration of knowledge systems. This technology not only effectively builds knowledge bases but also expands the scale of datasets. This research developed the Chinese medicine question-answering dataset CMQA20K, which contains 19,956 conversations covering over 200 scenarios applicable to most situations in daily life. Based on this dataset, an evaluation task for medical dialogue understanding and generation was designed, fine-tuning and testing the performance of the dataset on various open-source large language models. Experimental results indicate that the fine-tuned models outperform their non-fine-tuned counterparts in medical dialogue tasks. This study employs the LoRA (Low-Rank Adaptation) technique from Parameter-Efficient Fine-Tuning (PEFT) for model fine-tuning, effectively optimizing model performance through low-rank adaptation. The high-quality medical question-answering dataset constructed in this research provides an important foundation for future studies and applications. This dataset will soon be released in the GitHub project xuelangzhi163/TMedicalDialog.