Research And Design Of Knowledge Graph Based Question Answering System In Local Specialties

With the improvement of people's quality of life,attention to diet(especially local specialties)is also increasing.Our country is vast in size and abundant in resources,and each region has many local specialties with local characteristics.However,when people search for local specialty information in a certain region,the search engines always return a lot of repetitive content and only a more famous part of specialties is found.There are many specialties with good quality and high efficiency are not well known due to insufficient publicity or some other reasons.As a result,these specialties dull of sale.This article used natural language processing and knowledge graph technology to construct a knowledge graph(KG)in local specialties such as the efficiency and origin of local specialties,and designed and implemented a question answering system based on the KG.To some extent,this system can meet people's demand for healthy diet and alleviate the unsalable problem caused by the lack of circulation of specialty information.The main work of this article includes:(1)Constructed a knowledge graph of local specialties.A bottom-up method of KG construction is used.Firstly,we extracted information from the specialty website to obtain the entities and relationships of entities,and then aligned entities from different source.Finally,we organized and stored the KG by using the Neo4j database.(2)The question answering(QA)algorithm based on KG is designed and implemented.Aiming at the fact that there are only a few types of questions in this article,a template-based question analysis method is used.The question answering algorithm was divided into two tasks:named entity recognition and question classification.In the named entity recognition task,after comparing several models,the BERT-BiLSTM-CRF model which performed better was selected for the named entity recognition of questions.And the BiLSTM model with attention mechanism was used to realize question classification task.(3)Based on the KG and QA algorithm,a KGQA system of local specialties was designed and implemented which includes three parts:front-end display,database and QA model,and back-end data processing.This system can accurately answer the local specialty questions raised by users.The system is stable during the compression test and can support normal queries.