In the context of Semantic Search and Entity Retrieval, traditional approaches prepare an adhoc schema to index the contents of a Knowledge Graph and/or learn ranking functions out from annotated query sets (difficult and expensive to obtain, completely attached to a particular KG). To get rid of these limitations, we have devised a non-supervised approach based on information quantity measures which builds a template out from the KG to direclty use a regular indexing engine (e.g., ElasticSearch) and get equivalent results to supervised approaches regardless the domain. The results and the developed project were accepted at ESWC’25, held this year in Portoroz (Slovenia). #ESWC25
Congratulations to Samuel García! The paper can be found here.