Ontology concept matching method based on paired connected graphs and graph neural network

Abstract

The invention discloses an ontology concept matching method based on paired connected graphs and a graph neural network. Firstly, a pairwise connected graph of a source ontology and a target ontology is constructed, nodes of the graph are concept pairs, and edges of the graph correspond to attribute pairs; then, learning node embedding of the pairwise connected graph for predicting a matching relationship in the ontology; further, in order to obtain an ideal embedding result, automatically extracting similarity features from the attributes of the concept pairs by adopting a convolutional neural network; next, similarity features are propagated using a graph neural network and a final embedding of the concept pairs is obtained. And finally, according to a concept pair embedding result obtained by learning, predicting by a general classification model to obtain a concept matching result. According to the technical scheme, the problem of concept matching in ontology isomerism can be solved, and the method is easy to implement, high in precision and good in effect.

Description

technical field [0001] The invention belongs to the technical field of artificial intelligence, in particular to an ontology concept matching method based on a paired connected graph and a graph neural network. Background technique [0002] An ontology is an explicit and formalized specification of a shared conceptual model, defining concepts, concept hierarchies, and attributes of concepts, and using these concepts and attributes to capture related domain knowledge and provide a common understanding of the domain knowledge. However, when researchers associate and publish semantic data independently, differences in the construction process will lead to problems such as different names for the same concept and different value ranges, resulting in ontology heterogeneity. Ontology matching is an effective way to resolve ontology heterogeneity. [0003] Ontology matching techniques can generally be divided into term-based matching techniques, instance-based matching techniques ...

AI Technical Summary

Problems solved by technology

Existing structure-based ontology matching methods often only capture local structural information, such as the out-degree and in-degree of a node and its neighbor nodes, while ignoring the global structural information

In addition, the existing structure-based ontology matching methods are difficult to use the textual information of the entities in the ontology, and these defects will greatly affect the accuracy

Images