Semantic Similarity Calculation Based On Small-scale Knowledge Graph

The development of Semantic Web has led to new breakthroughs in many fields,such as semantic search,knowledge engineering,knowledge map,data connection,etc.The core of the Semantic Web lies in the expression and expression of the knowledge of the local layer,which involves relevant rules and reasoning.Semantic similarity is one of the main problems in these fields.Because of the large-scale,heterogeneous and loosely organized nature of Internet content,it challenges people to effectively access information and knowledge.Knowledge Graphics has a strong ability of open organization and semantic processing,which lays the foundation for the knowledge organization and intelligent application in Internet age.The current work on the main semantic similarity methods focuses on the structure of the semantic network between concepts(eg path length and depth),or on the conceptual information content(IC)only.In general,semantic similarity measures can be used to weight or rank similar concepts based on concept taxonomy.IC is a measure of the particularity of a concept.Higher values of the IC are associated with more specific concepts(eg,actor),while lower values are more common(eg,person).The IC is calculated based on the frequency count of the concepts that appear in the text corpora.The emergence of more specific concepts each time also means the emergence of more general ancestral concepts.In order to alleviate the disadvantages of path-based metrics and IC-based metrics,this paper proposes a new semantic similarity method,namely L-path.Combining these two methods,ICs are used to weight the shortest path length between concepts.In order to adapt the corpus-based IC method to the structured knowledge map,a graph-based IC calculation method is proposed in this paper.This method can make KGs-based semantic similarity measure based on KG without offline domain preparation.In the experiment,it is verified that the proposed method has a certain degree of feasibility and credibility in computing semantic similarity in knowledge graphs.Compared with other methods,the results are superior to other methods.