Research On Ontology Matching Based On Word Embedding And Structural Similarity

With the development of the digital information age,more and more information systems are used in different professional domains,which has also led to the continuous expansion of information capacity.How to effectively acquire,store,manage and apply these digital resources has become a problem that needs to be solved.The knowledge base based on ontology has emerged as the solution.Ontology is a conceptual model abstracted from the objective world,including concepts and the relationships between concepts.Ontology is the abstraction of knowledge and the formal expression of concepts and their relationships in a certain domain.It has the characteristics of conceptualization,formalization,explicit,and description of domain knowledge.In a specific domain,experts have different understandings of domain knowledge,or different purposes of constructing ontology.These will lead to multiple different ontologies in the domain.This phenomenon is called ontology heterogeneity.For the research fields that require cross-ontology operations such as knowledge fusion and knowledge reasoning,the ontology heterogeneity has caused certain difficulties for research.Ontology matching is one of the solutions to the ontology heterogeneity problem.It refers to the process of comparing the concepts in the ontology with each other and establishing a certain mapping relationship between them.For the text information in the ontology and the external information,the ontology matching model based on word embedding used in this thesis uses Siamese Continuous Bag-of-Words model with the knowledge distillation method to improve the word vector,so that it can distinguish between semantically similar and descriptively associated.Build contextual information for entities,and use BERT to generate dynamic word vectors for them to achieve the purpose of distinguishing polysemous words.Regarding the structural information in the ontology,the ontology matching model that introduces structural similarity used in this thesis converts the ontology into the structure of a graph and uses the Sim Rank algorithm to calculate the structural similarity between the entities in the graph,and verify the matching results,which is calculated by text information.In the matching process,the stable matching algorithm is used to obtain the stable one-to-one matching result,and at the same time,the oneto-many matching result is obtained by setting a similarity threshold,and compare the results of these two matching algorithms.All in all,this thesis aims at the problem of ontology heterogeneity,makes full use of the text information and structure information in ontology and external information,and proposes a new ontology matching model,which enriches and perfects the method system of ontology matching task.