Construction Of Question And Answer System For Repair Of Huizhou-style Architecture Based On Knowledge Graph

In recent years,China’s rapid urbanization has led to the destruction of traditional architecture to varying degrees.According to the third national cultural relics census,around 20,000 immovable cultural relics have been damaged due to urban construction,with roughly 5% of Hui-style architecture lost annually.The preservation and restoration of Huizhou-style architecture are influenced by a variety of factors,including natural environment,socio-economic development,local protection policies,and repair techniques.Unfortunately,the existing knowledge on the restoration of Huizhou-style architecture is primarily stored in books and traditional database,and is laden with technical terms and unclear boundaries.This makes accessing and utilizing the knowledge inefficient for restoration personnel.Additionally,traditional search engines are not able to accurately retrieve relevant information.To address this issue,it is necessary to extract,store,and manage knowledge on the restoration of Huizhou-style architecture in a structured manner.This would enable restoration personnel to access knowledge,thereby promoting the development of traditional architecture restoration and conservation technology towards intelligence.This paper presents an innovative approach based on ELECTRA for recogniting named entities and extracting knowledge from the restoration content of Huizhou-style architecture.This method effectively addresses issues related to unclear boundaries and complex types of restoration entities.To facilitate systematic and efficient management of knowledge,we utilize the Neo4 j graph database to construct a knowledge graph for the restoration of Huizhou-style architecture.Additionally,we propose a method for natural language queries of the knowledge graph,enabling users to access and utilize the information more efficiently.To provide restoration personnel with accurate and efficient question and answer services,we design and implement a knowledge graph-based question and answer system for emblematic building restoration.The main research work of this paper are summarized as follows:(1)Named entity recognition in the field of restoration of Huizhou-style architectureTo address the issue of the high number of technical terms and specialized vocabulary,as well as the complexity of entity types in the restoration of Huizhou-style architecture,an ELECTRA-based named entity recognition method is proposed.This method utilizes ELECTRA as the word embedding layer,employs Bi LSTM for context feature extraction for rare entities and specialized terms in the field,and incorporates multi-head self-attention mechanism to capture important information in different positions and representation subspaces,resulting in improved entity recognition accuracy.a CRF sequence label decoding optimizer is utilized to output the maximum probability sequence result.Comparative experiments on a self-built dataset were conducted among Bi LSTM-CRF,BERT-Bi LSTM-CRF,and BERT-Bi LSTM-MHA-CRF models.The proposed method achieved an F1 score of 91.07%,showing an improvement of 4.68%,4.42%,and 2.21%,respectively,compared to the other models.(2)Knowledge graph construction for the restoration of Huizhou-style architectureTo address the issue of diverse knowledge forms and inconsistent structures in the restoration of Huizhou-style architecture,this paper proposes to use knowledge graph technology to systematize fragmented knowledge and better establish connections between them.Knowledge extraction is performed on the obtained text data,and then knowledge reduction is used to solve problems of polysemy and ambiguity.The resulting knowledge is stored in CSV files in the form of triples,and then imported into a Neo4 j graph database using the LOAD CSV method.To address problem of diverse and uneven quality of user natural language queries,this paper proposes a topic entity matching method to identify the main entity in the question,and a BERT-Text CNN intent classification method to determine the intent of the question.By mapping the relationship chain to the knowledge graph,the natural language question can be fully understood,and a Cypher query statement can be generated to retrieve knowledge from the knowledge graph.(3)Design and implementation of a question and answer system for restoration of Huizhou-style architectureIn response to the lack of functions in existing knowledge graph-based Q&A services for Huizhou-style architecture restoration,this paper studies design and implementation of a knowledge graph-based Q&A system for Huizhou-style architecture restoration.The research is based on the Huizhou-style architecture restoration knowledge graph as the data foundation,with Django,Bootstrap,and ECharts as the development framework and front-end visualization tools,and knowledge Q&A as the core business.The system is built based on knowledge graph technology to provide accurate and effective restoration and preservation knowledge for restoration personnel,and to realize the protection and inheritance of Huizhou-style architecture and regional culture.Figure [43] table [12] reference [77]...