Enhancing data discovery with contextual pre-filtering

In this paper, entity contextual pre-filtering is proposed to refine dataset relevance assessment and streamline data discovery. Heterogeneous Graph Neural Networks are used to exploit the local context embedded within graph-based schemas. The proposed pre-filtering approach is versatile and does no...

ver descrição completa

Detalhes bibliográficos
Autores: Flores Herrera, Javier de Jesús|||0000-0002-2998-9962, Nadal Francesch, Sergi|||0000-0002-8565-952X, Romero Moral, Óscar|||0000-0001-6350-8328
Formato: artículo
Fecha de publicación:2025
País:España
Recursos:Universitat Politècnica de Catalunya (UPC)
Repositorio:UPCommons. Portal del coneixement obert de la UPC
Idioma:inglés
OAI Identifier:oai:upcommons.upc.edu:2117/449055
Acesso em linha:https://hdl.handle.net/2117/449055
https://dx.doi.org/10.1186/s40537-025-01312-5
Access Level:acceso abierto
Palavra-chave:Data discovery
Metadata enrichment
Heterogeneous graph neural networks
Àrees temàtiques de la UPC::Informàtica::Sistemes d'informació
Descrição
Resumo:In this paper, entity contextual pre-filtering is proposed to refine dataset relevance assessment and streamline data discovery. Heterogeneous Graph Neural Networks are used to exploit the local context embedded within graph-based schemas. The proposed pre-filtering approach is versatile and does not rely on any specific similarity metric, making it applicable to a wide range of data discovery methods. The proposed technique increases data discovery precision by reducing false positives and identifying significant data relationships. This method has been empirically validated across a variety of real-world datasets to improve data discovery efficiency and accuracy.