User-guided Repairing of Inconsistent Knowledge Bases - Archive ouverte HAL Access content directly
Conference Papers Year : 2018

User-guided Repairing of Inconsistent Knowledge Bases

Abdallah Arioua
  • Function : Author
  • PersonId : 961936
Angela Bonifati

Abstract

Repairing techniques for relational databases have leveraged in- tegrity constraints to detect and then resolve errors in the data. User guidance has started to be employed in this setting to avoid a prohibitory exploration of the search space of solutions. In this pa- per, we present a user-guided repairing technique for Knowledge Bases (KBs) enabling updates suggested by the users to resolve errors. KBs exhibit more expressive constraints with respect to relational tables, such as tuple-generating dependencies (TGDs) and negative rules (a form of denial constraints). We consider TGDs and a notable subset of denial constraints, named contra- diction detecting dependencies (CDDs). We propose user-guided polynomial-delay algorithms that ensure the repairing of the KB in the extreme cases of interaction among these two classes of constraints. To the best of our knowledge, such interaction is so far unexplored even in repairing methods for relational data. We prove the correctness of our algorithms and study their feasibility in practical settings. We conduct an extensive experimental study on synthetically generated KBs and a real-world inconsistent KB equipped with TGDs and CDDs. We show the practicality of our proposed interactive strategies by measuring the actual delay time and the number of questions required in our interactive framework.
Fichier principal
Vignette du fichier
paper-61.pdf (4.94 Mo) Télécharger le fichier
Origin Files produced by the author(s)
Loading...

Dates and versions

hal-01979680 , version 1 (14-01-2019)

Licence

Identifiers

Cite

Abdallah Arioua, Angela Bonifati. User-guided Repairing of Inconsistent Knowledge Bases. EDBT 2018 - 21st International Conference on Extending Database Technology, Mar 2018, Vienna, Austria. pp.133-144, ⟨10.5441/002/edbt.2018.13⟩. ⟨hal-01979680⟩
198 View
119 Download

Altmetric

Share

Gmail Mastodon Facebook X LinkedIn More