[Objective] This paper focuses on breaking through the difficult problem of automatic identification and extraction of patent technology and function entities, intelligently perceiving and generating key technology and function in patent documents, and assisting in the high-quality construction of patent technology and function matrix. [Methods] In this paper, we propose a new idea of ChatGPT applied to patent technical efficacy entity extraction task, and use the method of ChatGPT+Prompt to realize the recognition, extraction and generation of patent technology words, function words and technology-function binary groups.[Results]This paper recognizes and generates patent technology and function entities in four domains and three languages, and the experimental results (ROUGE values) comparing cross-domain, cross-language, and prompted sample sizes show that the method can recognize technology-function binary groups more accurately. By comparing the ROUGE-1 values, it can be seen that the new energy automobile field has the best effect, English patents have the best performance, the cross-domain and cross-linguistic abilities are significant, and giving one-shot will significantly improve the model performance.[Limitations]The method in this paper still has limitations such as the lack of standards for Prompt, the duplicity of generated content, and the choice of one-round or multi-round Q&A.[Conclusions] The method in this paper still has problems such as the lack of standards for Prompt, the duplicity of generated content, and the choice of one-round or multi-round Q&A. [Conclusion] The method proposed in this paper possesses rationality and feasibility, effectively reduces the labor cost and task threshold of technology and function entity generation, expands the application scenarios of AIGC, and releases the great potential of ChatGPT in patent document mining. At the same time, we put forward thoughts and suggestions for users to understand ChatGPT to assist in carrying out generation-type tasks.