Design And Implementation Of Wheat Variety Question Answering System Based On Knowledge Graph

In order to solve the problems existing in the traditional Internet agricultural knowledge question answering system,such as classification of short questions,low accuracy of user intention understanding and matching,and difficult to guarantee the quality of answers,this paper takes wheat variety knowledge question answering as an example,and according to domain knowledge ontology technology,collects wheat variety data and short questions data set,and constructs wheat variety Knowledge Graph.Using natural language processing and machine learning techniques,the task of user intention understanding of wheat varieties short questions was transformed into short question classification and question template matching.By using a variety of machine learning and natural language processing techniques,the classification model of short questions and question template matching model were proposed and optimized.The accuracy and effectiveness of short questions classification and question answering of wheat varieties were ensured through the construction of user intention recognition model.Finally,the optimal wheat variety short questions template matching model was selected for integrated development,and a wheat variety question answering prototype system based on knowledge graph was designed and implemented.The main research results and conclusions are as follows:1.Construction of wheat variety Knowledge Graph based on ontology.Wheat variety knowledge belongs to vertical domain knowledge.In order to improve the systematicness,completeness and expansibility of the ontology,this paper adopts the top-down method to construct the knowledge system of wheat variety,so as to meet the demands of users at different levels and with different needs for wheat variety knowledge question-answering service,and designs four levels of wheat variety knowledge ontology.Established the field and range of wheat variety knowledge.At the same time,for wheat variety data collection,11151 wheat variety approval information was crawled from China Crop germplasm Information Network and seed business network.On the basis of data crawling,cleaning,extraction and fusion,18404 wheat variety entities and 131653 relationships between entities were designed and extracted,and the wheat variety Knowledge Graph based on RDF was established.Provide data preparation for intelligent question and answer.2.Design and verification of wheat variety question answering model.Through domain experts and rule-based template construction,9 kinds of question templates are designed,142811 short questions and 116 types of natural language questions are constructed.In order to recognize the named entity of wheat varieties,Han LP was used to recognize the named entity of short questions,and a wheat variety dictionary was constructed.The user intention of short questions was divided into the matching task of short question entity recognition and question template.Based on machine learning methods such as naive Bayes(NBC),bi-coded Representation quantity(BERT)based on converter and Bi-LSTM+Attention(Bi-LSTM+Attention),a wheat variety question answering model was designed and validated.Through the training and testing of "7:3" data sample ratio,we evaluated the performance of three machine learning models,NBC,BERT and Bi-LSTM+Attention,on the classification matching of short questions and question templates.The results showed that the accuracy rates of the three models were 92.85%,94.45% and 96.59% respectively,and all of them could provide high-quality classification matching results.Bi-LSTM+Attention has the best performance,and its accuracy rate,recall rate and F1 value are all higher than 96.4%,which is better than the other two methods.The experiment shows that the combination of machine learning model and knowledge graph can realize the short question to question template matching of wheat varieties with high quality,which can meet the user’s need of intention recognition and provide quality assurance for accurate question answering.3.Wheat variety question-and-answer prototype system based on lightweight framework.Based on wheat variety Knowledge Graph,integrated named entity recognition technology,user intent matching model based on Bi-LSTM+Attention and database retrieval technology based on Neo4 j map,and adopted lightweight Spring Boot framework at the back end.The front-end user page uses HTML,CSS,Java Script language and Thymeleaf page rendering technology to design and implement the wheat variety question answering system based on knowledge graph,which has the function of information retrieval,intelligent question answering and so on,and has important value and significance for the development and popularization of wheat variety knowledge services.