Article Dans Une Revue Lecture Notes in Business Information Processing Année : 2025

Alignment of Schema-Only and Instance-Only Data Sources Using Large Language Models

Résumé

Data matching is a persistent challenge in heterogeneous data integration. Traditional data matching methods, relying on schemabased, instance-based, or hybrid approaches, often fall short when aligning disparate data from schema-only with instance-only sources. To address this problem of aligning disparate sources, we present an innovative data matching framework that enables the matching of one data source, where only schema-related information is available, with another data source, where only instance-related information is available. The strength of our framework lies in its ability to combine outputs from multiple Schema-Instances matchers, generating auxiliary information to enhance the alignment between disparate data structures. Our framework is validated using the Valentine Benchmark through extensive experiments. These findings underscore the potential of our approach to advance the integration of diverse data sources.

Fichier principal
Vignette du fichier
RCIS_2025_paper_62.pdf (775.95 Ko) Télécharger le fichier
Origine Fichiers produits par l'(les) auteur(s)
Licence

Dates et versions

hal-05091439 , version 1 (31-05-2025)

Licence

Identifiants

Citer

Nour Elhouda Kired, Franck Ravat, Jiefu Song, Olivier Teste. Alignment of Schema-Only and Instance-Only Data Sources Using Large Language Models. Lecture Notes in Business Information Processing, 2025, pp.67 - 85. ⟨10.1007/978-3-031-92471-2_5⟩. ⟨hal-05091439⟩
2081 Consultations
157 Téléchargements

Altmetric

Partager

  • More