An agent system for information retrieval in an academic environment

• Main Author: Khalifa Chekima
• Format: Thesis
• Language: English; English
• Published: 2012
• Subjects: QA76.75-76.765 Computer software
• Online Access: https://eprints.ums.edu.my/id/eprint/41797/1/24%20PAGES.pdf; https://eprints.ums.edu.my/id/eprint/41797/2/FULLTEXT.pdf; https://eprints.ums.edu.my/id/eprint/41797/

In the recent years, academic domains just like other domains have undergone a tremendous growth on both content and users. This growth has lead to information overload in which academician are finding it difficult to locate the right academic paper at the right time. Search Engines, were designed originally to be helpful in searching for relevent resluts and returning them to users. Yet, due to thousands of petentially relevant sites, thus search engines are losing their usefulness. To address this problem, a development of a reliable multi-agent system that is able to guide academician through the big ocean of information by filtering the information and recommending them relevant papers is vital. However, recommending an item/academic paper is not easy since it depends on many factors such as the user’s current interest, as user’s interest changes over time, size of content, and number of users. This thesis presents the development of multi-agent system that helps academicians in the process of retrieving relevant academic papers by recommending them papers based on their current interest. The recommendation is generated using a Hybrid recommendation approach, which is a combination of the two well known recommendation approach, content-based filtering approach and collaborative filtering approach. The system consists of four agents working together. The first agent is Monitoring Agent that monitors User’s browsing behavior to implicitly observe users’ current interest. The second module is the Categorizer Agent that automatically organizes papers downloaded by users into subcategories based on ACM Association Computing Machinery CCS (Computing Classification System) structure by considering papers’ content similarity. The third agent is the Recommender Agent that recommends papers to users based on Hybrid approach, and the last agent is the Search Agent that allows users to search for academic papers locally. The use of multi-agent technology has overcome many problems that a traditional recommendation system suffers from. The accuracy of the Hybrid approach used by the proposed system in the recommendation is then compared with two other common recommendation approaches, content-based filtering approach and collaborative filtering approach by counting the precision value of each approach. Based on the results, the system was able to recommend well based on user’s current interest using Hybrid approach. Besides that, the categorizer agent has shown promising results in categorizing of academic papers based on the proposed ACM CCS system.