The Design And Implementation Of An Intelligent Retrieval System Based On Semantic Mapping

Enterprise documents are important digital assets,which record the domain knowledge accumulated in the enterprise.Today,the digital transformation of enterprises has become a mainstream.The management and retrieval of enterprise documents are becoming increasingly complex.Exploiting software engineering technology to manage and retrieve the knowledge in enterprise documents can make great benefits in improving the work efficiency of each team in an enterprise.By combining the actual situation in the enterprise and extracting the information within the enterprise documents and work-order tickets,this thesis uses knowledge graph to improve the document storage mode,designs a new retrieval method based on semantic mapping,and finally realizes an intelligent enterprise document knowledge retrieval system,which improves the work efficiency between different teams.The main work of this thesis is as follows:1)A document storage mode is built based on knowledge graphs.This thesis analyzes the storage structure of documents in an enterprise,summarizes some notable features,such as a large number of enterprise documents,fine content of a single document,wide application of document labels,much domain-specific content.This thesis puts forward a method of storing enterprise documents based on the knowledge graph model.With the help of natural language processing technology,a model is designed for building knowledge graphs automatically.With the technologies of automatic labeling and data enhancement,the problem of lacking large-scale domain dataset is resolved.After comparing the target applicability of different storage approaches,the method Elastic Search is finally selected to realize the storage and retrieval of knowledge graphs.2)A document retrieval method is designed based on semantic mapping.By making use of knowledge graphs for data storage,this thesis improves the traditional retrieval method of knowledge graphs and proposes a retrieval method based on the three-time semantic mapping,which can effectively map natural language questions into entities in the knowledge graph and has higher precision rate than the traditional method.When the candidate entities be obtained,a template-based query statement construction method is then designed to retrieve entities in knowledge graph.Finally,a full-text retrieval mechanism is designed to improve the recall rate of the traditional method.Experiments shows the effectiveness of above methods in retrieving documents,which have higher precision rate and recall rate than the traditional retrieval methods of knowledge graphs.3)The intelligent retrieval system is implemented.This thesis conducts the demand investigation and analysis of the development team and after-sales support team in the enterprise,respectively.Based on the actual requirements,the above retrieval method is applied.An intelligent retrieval system is designed and implemented,which is suitable for retrieving enterprise document knowledge database.By using advanced front-end and back-end technologies,the system is deployed on a cloud server.The automatic testing and manual testing both show the user-friendliness and effectiveness of the intelligent retrieval system.Compared with the traditional retrieval system,its retrieval performance in keyword retrieval and natural language retrieval has been improved.The designed system has already been put into test within the enterprise and obtained positive feedback.It shows that the Intelligent retrieval system can effectively improve the work efficiency between different teams.The design idea introduced in this thesis can also help other enterprises to improve their document knowledge management and retrieval process,which has high practical significance in improving the workflow.