PhD Thesis of Frederic Dang Ngoc
My PhD Thesis in CS was entitled "Decentralised and Personalised Search Engine"
and was supervised by Pierre Fraigniaud (LRI, University of Paris-Sud, Orsay) and Joaquin Keller (France Telecom R&D Division).
I defended my PhD on February 17th, 2006 at the University of Paris-Sud.
Thesis commitee members
- Chair and Examiner: Marie-Christine Rousset, Professor, University of Grenoble
- Co-Advisor: Joaquín Keller, R&D Engineer, France Telecom R&D Division
- Advisor: Pierre Fraigniaud, Research Director CNRS, LRI, Universiy of Paris-Sud
- Reviewer: Karl Aberer, Professor, Ecole Polytechnique Federale de Lausanne(EPFL)
- Reviewer: Anne-Marie Kermarrec, Research Director INRIA, IRISA
- Examiner: Ernst W. Biersack, Professor, Eurecom
Abstract
Nowadays, information retrieval is still a hard problem. The issue does not lie on information scarcity but rather on its overabundance. In this thesis, we will describe the search engine MAAY that we have conceived. The motivation is to propose an alternative to web search engines (Google, Yahoo !, etc.) whose lack of transparency makes less credible (skewed results, censorship, privacy, etc.). In this system, we propose to use the users, more particularly, their actions in the system, to personalize results. In MAAY, publications, indexations and searches are decentralized on user’s computers. Each computer (peer) learns semantic profiles of other peers and of documents from observations of search and download messages. By preferentially querying peers with similar semantic profiles, and by ranking documents according to the requester profile, users may find documents which are relevant for them. That is what we show through simulations. To do so, we propose a user and document model, and several measures to evaluate the efficiency of a search system to provide personalized answers to its users. The results that we have obtained show that MAAY succeeds in adapting to its users by providing them personalized answers. Finally, we have implemented a prototype of our system to test its feasibility and the functionalities that will be implemented in the final version of our system.
Keywords: Search Engine, Peer-To-Peer
