Study On Cultural Relics Question Answering System Based On Knowledge Graph

With the rapid development of the Internet,more and more museums in China display cultural relics on the network platform,and more and more data resources are available on the Internet.Facing the rapid growth of information,it is more and more difficult for users to accurately obtain information.At present,people mainly obtain the information they need through search engines,but the information returned to users in this way is often redundant,requiring users to filter useless information by themselves,which brings great inconvenience and takes up time.Therefore,many technology companies and related research institutions have invested a lot of time and energy in the research and development of intelligent question-answering robots,of which the core module is the question-answering system.However,in the field of cultural relics,it is difficult for people to find the required knowledge due to the large and complex knowledge.Therefore,this paper mainly studies the question answering system based on cultural relic knowledge map.In order to facilitate the search of cultural relics related knowledge.This paper first analyzes and introduces the relevant knowledge and key technologies in the field of question and answer,as well as the construction process of knowledge graph.In this study,knowledge triad is extracted from cultural relic data to construct cultural relic knowledge map.On the basis of cultural relic knowledge map,question answering system is studied.The design of question answering system mainly includes several modules: entity identification,question matching and answer finding.The main research contents are as follows:(1)Transformer combined with the network model of BILSTM-CRF is used to extract named entities in questions,and position information is added to the input vector.At the same time,BILSTM network is used to extract direction information to improve the accuracy of entity recognition.(2)TF-IDF algorithm with word order fusion was used to match the problem template.The preprocessed question is abstracted into template form to match with question template set,and the word order is fused to calculate the similarity of TF-IDF sentence vector.(3)Find and return the answer.By using the CYPHER statement corresponding to the problem template and combining with the identified entity,the answer is queried in the Neo4 j graph database and returned.Based on the above analysis and research,the research and design of the question and answer system based on the map of cultural relic knowledge is completed.Finally,the test results show that the accuracy of the answer is higher than 85%.The experimental results show that the cultural relic question and answer system in this paper can help users query related questions about cultural relic.