Similarity Measures For Incomplete Database Instances - Archive ouverte HAL Accéder directement au contenu
Communication Dans Un Congrès Année : 2024

Similarity Measures For Incomplete Database Instances

Boris Glavic
  • Fonction : Auteur
  • PersonId : 1371029
Renée J Miller
  • Fonction : Auteur
  • PersonId : 1371030
Paolo Papotti
Donatello Santoro
  • Fonction : Auteur
  • PersonId : 1296826
Enzo Veltri
  • Fonction : Auteur
  • PersonId : 1296825

Résumé

The problem of comparing database instances with incompleteness is prevalent in applications such as analyzing how a dataset has evolved over time (e.g., data versioning), evaluating data cleaning solutions (e.g., compare an instance produced by a data repair algorithm against a gold standard), or comparing solutions generated by data exchange systems (e.g., universal vs core solutions). In this work, we propose a framework for computing similarity of instances with labeled nulls, even of those without primary keys. As a side-effect, the similarity score computation returns a mapping between the instances’ tuples, which explains the score. We demonstrate that computing the similarity of two incomplete instances is NP-hard in the instance size in general. To be able to compare instances of realistic size, we present an approximate PTIME algorithm for instance comparison. Experimental results demonstrate that the approximate algorithm is up to three orders of magnitude faster than an exact algorithm for the computation of the similarity score, while the difference between approximate and exact scores is always smaller than 1%.
Fichier principal
Vignette du fichier
compareEdbt24.pdf (891.55 Ko) Télécharger le fichier
Origine : Fichiers produits par l'(les) auteur(s)

Dates et versions

hal-04531944 , version 1 (04-04-2024)

Identifiants

  • HAL Id : hal-04531944 , version 1

Citer

Boris Glavic, Giansalvatore Mecca, Renée J Miller, Paolo Papotti, Donatello Santoro, et al.. Similarity Measures For Incomplete Database Instances. EDBT 2024, 27th International Conference on Extending Database Technology, IEEE, Mar 2024, Paestum, Italy. ⟨hal-04531944⟩

Collections

EURECOM ANR
0 Consultations
0 Téléchargements

Partager

Gmail Facebook X LinkedIn More