A Semantic Understanding Method for Patent Text Based on Large Language Model
Chenchen Zhang, Jian Sun, Chao Luo, Fan Chen, Bo Li, Luo Chen · 2023
The large language model has powerful natural language understanding and generation capabilities, which can automatically extract keywords and key information from literature, thus achieving fast and efficient patent retrieval. At the same time, the large model can also be customized according to user needs, improving retrieval accuracy and efficiency. A semantic understanding method for patent text based on large language model is proposed in this paper. This method first requires preprocessing of the dataset, inputting the problem into ChatGPT according to the specified instruction format, and then classifying the output content for analysis based on different classifications. Then, the entity recognition results are input into the patent system, and the knowledge graph in the patent system is used for inference analysis to obtain the corresponding answer to the problem. The research results show that the method based on the large language model has high accuracy and generalization ability, and can achieve semantic understanding and analysis of patent texts.