MOR : Mesure orientée rappel pour les systèmes de recherche d’information

Résumé : Most evaluation metrics in Information Retrieval (IR) give more weight to precision, which is consistent within several contexts, such as web retrieval, where users check only few documents appearing at the beginning of the result lists. These metrics are not able to measure the satisfaction of the user in contexts where the IR system is supposed to find the maximum of relevant documents, which is the case of patent or medical retrieval. In this article, we define the basic elements that should be considered to evaluate and compare IR systems correctly in a recall-oriented context. We demonstrate that previous measures do not consider the totality of these elements. We propose MOR which is a recall oriented measure that allows the evaluation and the comparison of information retrieval systems in a recall context.
Type de document :
Article dans une revue
Revue des Sciences et Technologies de l'Information - Série Document Numérique, Lavoisier, 2015, 18 (1), pp 37 - 54. 〈http://www.cairn.info/revue-document-numerique.htm〉
Liste complète des métadonnées

https://hal-emse.ccsd.cnrs.fr/emse-01158893
Contributeur : Florent Breuil <>
Soumis le : mardi 2 juin 2015 - 09:47:35
Dernière modification le : lundi 21 mars 2016 - 17:43:45

Identifiants

  • HAL Id : emse-01158893, version 1

Citation

Bissan Audeh, Philippe Beaune, Michel Beigbeder. MOR : Mesure orientée rappel pour les systèmes de recherche d’information. Revue des Sciences et Technologies de l'Information - Série Document Numérique, Lavoisier, 2015, 18 (1), pp 37 - 54. 〈http://www.cairn.info/revue-document-numerique.htm〉. 〈emse-01158893〉

Partager

Métriques

Consultations de la notice

86