Design And Implementation Of Chinese Recipe Intelligent Question Answering System Based On Knowledge Grap

As society continues to progress,diet-related information has become a major concern in people’s daily lives.Although the development of Internet technology has made it easier for people to obtain information related to recipes,such information is often simply aggregated without filtering,making it difficult to meet people’s requirements for information refinement.The rise and development of knowledge graphs provide a solid foundation for the construction of intelligent question and answer systems and high-quality databases.Based on this,this paper designs and develops an intelligent question and answer system for Chinese recipes based on the knowledge graph of Chinese recipes.The research work in this paper consists of the following:(1)Research on the knowledge graph construction method of Chinese recipes.To address the characteristics of Chinese recipes with multiple sources and heterogeneity,a top-down approach is adopted to construct the Chinese recipe knowledge graph.The ontology concept layer of the Chinese recipe knowledge graph is first designed,and then the extracted triples are added to the data layer according to the ontology specification of the concept layer.To address the problem of multiple and overlapping triads in the original data,this paper proposes a joint entity relationship extraction model,BERTAdvCasLSTM,based on adversarial training and pointer cascade annotation.The model uses BERT to train word vectors and extracts the head entities in triads and the tail entities under specific relationships by means of pointer cascade annotation,solving the extraction problem of multiple and overlapping triads and using adversarial training to improve the generalization ability of the model.To better extract tail entities,this paper uses BiLSTM to fuse the head entity features with the sentence features.The experimental results have shown that the model proposed in this paper outperformed the CasRel model in terms of F1-score,with improvements of 2.7% and 1.9% observed in the DuIE1.0 and Recipe＿Chinese datasets,respectively.These results demonstrate the effectiveness of our proposed model.(2)Research on intelligent question and answer methods for Chinese recipes based on knowledge graphs.For the annotation characteristics of Chinese recipe data,an interrogative entity recognition model based on GlobalPointer and adversarial training is designed to recognize entities in interrogative sentences,solving the recognition problem of nested entities.Aiming at the short length of the user’s natural language questions and various expression forms,the question intent understanding model based on BERT-TextCNN is adopted to understand the intent of the user’s questions.Construct a Cyhper query statement based on the obtained information to complete the answer query.The experimental results show that compared with the BERT-CRF model,the F1-score of the interrogative named entity recognition model has improved by 1.77%on the Recipe＿NER dataset.Compared with the BERT model,the F1-score of the interrogative intent recognition model on the Chinese recipe interrogative intent recognition dataset has improved by 1.08%.(3)Design and implementation of a knowledge graph-based intelligent question and answer system for Chinese recipes.This paper is based on the knowledge map of Chinese recipes,uses the Django development framework,takes the Chinese recipe question and answer method based on the knowledge map as the technical route,uses the Echarts visual chart tool to visualize the knowledge map,and builds an intelligent question-answering system for Chinese recipes based on the knowledge map.The system can help users acquire Chinese recipe knowledge efficiently and accurately and promote the digital and intelligent development of Chinese recipe knowledge.