Skip to Main content Skip to Navigation

Vers des systèmes de recommandation robustes pour la navigation Web : inspiration de la modélisation statistique du langage

Geoffray Bonnin 1
1 KIWI - Knowledge Information and Web Intelligence
LORIA - Laboratoire Lorrain de Recherche en Informatique et ses Applications
Abstract : The goal of this thesis is to enhance the quality of the recommender systems for Web navigation by using sequentiality. The notion of sequentiality has been widely studied in the frame of Web recommendation. Such studies usually attempt to provide a trade-off between accuracy, space and time complexity, and coverage. Two extra problems are the presence of noise within the sessions of navigations (navigation mistakes, pop-ups, etc.), and parallel browsing. Most of the models that have been proposed in the literature either exploit low size contiguous sequences and are not robust to noise, or discontiguous sequences and induce large time and space complexities. This last problem can be lowered by performing a selection of sequences in order to lower space complexity, but this results in a coverage problem. Last, to the best of our knowledge, parallel browsing has never been studied in the frame of recommendation. The challenge of this thesis is thus to propose new sequential algorithms that have the five following characteristics: (1) a good precision of recommendations, (2) a good robustness to noise, (3) the ability to take into account parallel browsing, (4) a high coverage and (5) a low time and space complexity. In order to complete this challenge, we take inspiration from Statistical Language Modeling (SLM). Indeed, the general characteristics of Web navigation are very close to those of natural language. SLM dates back to a longer time than recommender systems and have widely proved their accuracy and efficiency. Moreover, most of the statistical language models that have been proposed take into account sequences. We thus investigated the exploitation of models used in language modeling and adapted them to the specific constraints of Web navigation.
Document type :
Complete list of metadatas

Cited literature [149 references]  Display  Hide  Download
Contributor : Geoffray Bonnin <>
Submitted on : Friday, April 1, 2011 - 4:20:10 PM
Last modification on : Monday, October 19, 2020 - 11:00:02 AM
Long-term archiving on: : Saturday, July 2, 2011 - 2:29:13 AM


  • HAL Id : tel-00581331, version 1



Geoffray Bonnin. Vers des systèmes de recommandation robustes pour la navigation Web : inspiration de la modélisation statistique du langage. Informatique [cs]. Université Nancy II, 2010. Français. ⟨tel-00581331⟩



Record views


Files downloads