Research And Implementation Of Entity Alignment Based On Knowledge Representation Learning Method

With the rapid development of the Internet,various types of information data have also exploded.These massive pieces of fragmented data,as important information resources,have been organized into structured knowledge data and managed in the form of knowledge graphs,which can be widely used.In the fields of semantic intelligent search,knowledge answering,and knowledge reasoning.Due to the openness of the Internet world,many institutions or organizations will build knowledge graphs according to their own needs and ideas.Then different knowledge graphs will be full of diversity and heterogeneity,and there will be many semantically overlapping entities or data in their data.relationship.If multiple knowledge graphs are to be correlated with each other to build a larger-scale knowledge graph,knowledge fusion must be performed.The key technology is entity alignment.There are a lot of deficiencies in the traditional entity alignment tools.The application range of the entity alignment algorithms provided is very limited,and the efficiency is low,which can not meet the user's diverse task requirements;and the lack of a friendly human-computer interaction interface makes the use threshold of the tool relatively High,poor user experience.Aiming at the shortcomings of traditional entity alignment tools,this paper studies the entity alignment method based on knowledge representation learning and finds that it is not limited by linguistic features and structural information,and can obtain the semantic information inherent in text without relying on string information.A cross-knowledge graph entity alignment algorithm based on neural tensor network was proposed,and a knowledge graph entity alignment system was developed and implemented.The main research results of this paper include:(1)A cross-knowledge graph entity alignment algorithm based on neural tensor network is proposed,which is divided into joint knowledge representation learning and improved NTN alignment model.The joint knowledge representation learning method is based on the seed set for knowledge representation learning of multiple knowledge graphs.The alignment relationship of the seed set is regarded as a special relationship between entities as a constraint on the merger of two knowledge graphs,so that the two knowledge graphs are mixed in Together,the knowledge representation learning method is used to implement the vectorized representation of the two knowledge graphs in a unified low-dimensional space;the improved NTN alignment model is used to calculate the similarity between the vectors of the entities to be aligned,thereby inferring and judging the two entities to be aligned Whether there is an alignment relationship between pairs to achieve cross-language knowledge graph entity alignment.The experimental results of the method in this paper on the DBP15k dataset have a Hit@10 index of up to 79.20 and an MRR index of 0.511.The results show that the algorithm has better performance than traditional algorithms.(2)Design and implement a knowledge graph entity alignment system,which encapsulates the cross-knowledge graph entity alignment algorithm based on neural tensor network proposed in this paper,and uses it for online entity alignment calculation tasks and reserves an interface for other The extension of the entity alignment algorithm realizes efficient data caching and storage of knowledge map data,and provides a good human-computer interaction interface.Tests show that the knowledge graph entity alignment system generally meets the system's design goals,and has good stability,user-friendliness,efficiency,and flexibility.