원문정보
초록
영어
Merging query-hits in large scale systems, like P2P, is challenging and potentially complex because results have to be ranked with respect to each other while sources are heterogeneous and with no centralized control. To solve this problem, we advocated in [10] a knowledge-based approach relying on users’ profiles. A user profile includes information about past interests derived from the user past actions as well as information about peers from which results were obtained in the past for similar queries. Using a knowledge base can lead to the system obsolescence unless an effective approach is proposed to evolve this learned knowledge. Most used approaches for knowledge update are periodic and cannot react on user needs changes at the appropriate time. For this reason, we propose, in this paper, a controlled and distributed mechanism for knowledge evolution based on need observers. A need detector aims to detect the user new needs expressed in his queries, as well as the new resources. Experiments show a clear improvement of the system performance with our controlled mechanism.
목차
1. Introduction
2. Problem Statement
3. Related Works
4. A Controlled Knowledge Evolution Mechanism
4.1. Overview
4.2. Proposed Architecture
4.3. Updating Process
5. Experiments
5.1. Simulation Environment and Parameters
5.2. Test Scenarios
5.3. Experimental Results
6. Conclusion
References