A Rule-Based Multidimensional Axiomatic Fuzzy Set Knowledge Graph Question-Answering Model

The pretrained large language model has made significant progress in intelligent question-answering, reading comprehension, and other aspects. This paper refers to a rule-based mathematical model-the multidimensional axiomatic fuzzy set (AFS) model and extends it by using logic inverse and fuzzy implication operations to increase the interpretability of this model. For open-book question answering (QA), AFS theory can generate a knowledge graph of passages, and retrieve the context most closely related to the question on the basis of the graph retrieval algorithm. We utilize the reading comprehension ability of the large language model to analyze the retrieved context on the basis of the question and obtain the final answer. The experimental results show that the proposed model can perform better on several datasets.