Designing And Developing A System For Building And Querying Knowledge Graph

As an information management tool,knowledge graph can not only effectively handle large-scale data and semantic information,but also provide high-quality data presentation and display capabilities.This project designs and develops a system for building and querying knowledge graphs aimed at weapons,fully leveraging the application advantages of knowledge graph in data processing,and assisting in displaying the associations between different weapon entities through visualization,providing a more efficient and accurate data management solution for the intelligent construction of weapons.This article is structured around three key points:1.Knowledge joint extraction algorithm based on Seq-to-Seq framework and RoBERTa modelThis article constructs a raw dataset in the weapon field and designs its domain ontology to offer pattern-level information for knowledge extraction.To solve the problem of one main entity corresponding to multiple dependent entities in the dataset,this article designs a knowledge extraction model based on the Seq-to-Seq framework and combined with the RoBERTa model.The model integrates entity recognition and relationship classification through a sequence labeling task that prioritizes the main entity,using a semi-pointer-semi-label sigmoid structure.The loss function is also optimized with a power function,and the model has been verified to achieve good results.2.Knowledge graph retrieval model based on node matching and query expansionThis article proposes a knowledge graph retrieval model for weapons based on node matching and query expansion.The model first preprocesses user search terms through synonym conversion,entity name indexing,and fuzzy matching,so that they can be mapped to specific nodes in the knowledge graph.Then,using calculation methods such as edit distance,attribute matching,and Jaccard coefficient,semantic similarity calculations are divided into two categories and used in different node matching result scenarios,thereby achieving semantic understanding of user search keywords and query expansion.Experimental results have shown that this model improves the single matching mode of existing retrieval schemes and can return results that better meet user needs.3.Knowledge graph build and query system for the field of weaponsIn order to better utilize knowledge about weapons,this article integrates the ontology construction method and entity relation joint extraction model mentioned in the previous sections and designs and implements a knowledge graph construction and retrieval system for the weapon field.This system consists of four layers:data,model,business presentation.In terms of functionality design,it mainly includes six modules:data processing,ontology construction,knowledge extraction,knowledge storage,knowledge retrieval,and visualization interaction.This system provides users with a concise and unified way of data management and further realizes the efficient utilization of the weapon knowledge graph.