Chinese Text Open Domain Tag Generation Method via Large Language Model
Chunhui He, Bin Ge, Chong Zhang · 2024
The task of Chinese open domain tag generation is a difficult problem in the field of natural language processing (NLP). Considering that the existing methods lack the ability of induction and reasoning on the open domain tag generation task. In order to solve the above problems, this paper proposes a Chinese open domain tag generation method combined with large language model (LLM). Its core idea is to use large language model combined with prompt learning technology to design high-quality prompt schema to induce and reason the original text, so as to give corresponding inductive reasoning results. Then, the inductive reasoning results are processed by combining tag filtering and sorting algorithm to generate the open domain text tags. The experimental results show that the comprehensive performance of the proposed method is better than all baseline methods when generating Top-3 or Top-6 tags, and the accuracy can reach 34.3% and 39.6% respectively.