Large Language Models (LLMs) have demonstrated remarkable success in diverse natural language processing (NLP) tasks in general domains. However, LLMs sometimes generate responses with the hallucination about medical facts due to limited domain knowledge. Such shortcomings pose potential risks in the utilization of LLMs within medical contexts. To address this challenge, we propose knowledge-tuning, which leverages structured medical knowledge bases for the LLMs to grasp domain knowledge efficiently and facilitate trustworthy response generation. We also release cMedKnowQA, a Chinese medical knowledge question-answering dataset constructed from medical knowledge bases to assess the medical knowledge proficiency of LLMs. Experimental results show that the LLMs which are knowledge-tuned with cMedKnowQA can exhibit higher levels of accuracy in response generation compared with vanilla instruction-tuning and offer a new trustworthy way for the domain adaptation of LLMs. We release our code and data at https://github.com/SCIR-HI/Huatuo-Llama-Med-Chinese .
Named entity recognition (also known as entity recognition, entity segmentation and entity extraction) is a sub task of information extraction. It aims to locate and classify named entities in text into predefined categories, such as people, organization, location, time expression, etc. Compared with English, there are more unsolved problems in Chinese named entity recognition. Named entities in English have obvious formal signs, that is, the first letter of every word in entities should be capitalized, and entity boundary recognition is relatively easy. Compared with English, the task of Chinese named entity recognition is more complex, and the recognition of entity boundary is more difficult. In this paper, we propose a named entity method by adding the word position, which embeds the word position of each word into the word vector, in order to better recognize the boundary of Chinese named entity. The experimental results show that the F1 value of the named entity recognition method proposed in this paper increases by about 1%.