Résumé: Multi-example query over knowledge graphs allows users to provide multiple example fragments to express their intention, overcoming the shortage of single example query asks users to give a specific complete query example. Unfortunately, the existing multi-example query ignores the semantic relevance of entities, and query results are inconsistent in compactness, which causes it difficult for users to check their interesting results. Therefore, we first define a multi-example query over ontology-label knowledge graphs to facilitate users to express their query intentions and improve the semantic relevance of query results. Secondly, we propose a matching pattern to analyze users' query intention to prioritize the more compact query results to improve users' satisfaction. Specifically, we first use the OELIndex to filter the search space and improve query efficiency. Then, we construct a connected matching pattern according to query examples to get a more compact query result. After that, we select the pattern fragments with the smallest amount for priority matching by assessing the number of candidate sets, significantly reducing the isomorphic computations. Finally, extensive experiments are carried out on real data sets to verify the effectiveness and practicability of our proposed method.

Langue: Anglais