Tailoring aggregated search per class of query

Arlind Kopliku (1) , Karen Pinel-Sauvagnat (1) , Mohand Boughanem (1, 2)

1 IRIT-SIG - Systèmes d’Informations Généralisées
2 UT3 - Université Toulouse III - Paul Sabatier

In this paper, we study the use of class-vertical preferences (cvp) for aggregated search (AS). Our aim is multiple: a) support the vertical selection process when queries include named entities, b) fill the vacuum in research of available features for AS. Our studies show that class-vertical preferences (cvp) are reliable. Different taxonomies are analyzed. By comparing 5 systems, we show that class-level preferences can improve significantly web search: we double recall without introducing false positives and unimportant results.

Mots clés: aggregated search; class preferences

Domaines: Informatique [cs]

Arlind Kopliku, Karen Pinel-Sauvagnat, Mohand Boughanem. Tailoring aggregated search per class of query. [Research Report] IRIT : Institut de Recherche en Informatique de Toulouse. 2015. ⟨hal-03198268⟩