An Agent-based Federated Learning Object Search Service

Introduction The Brazilian Ministry of Education provides free digital pedagogical content by means of the Virtual and Interactive Net for Education program (RIVED, 2009), distributing these objects through the International Base of Educational Objects repository (BIOE, 2010). The main goal of these programs is to aid in the development and distribution of electronic educational material by using Learning Objects (LO) as the foremost technology to publish and disseminate such material. The material is formed by educational activities, which may contain multimedia resources, animations, and simulations. To locate a particular object in a repository is a difficult problem depending on the rightful indexation and cataloging of its material. This process corresponds to the fulfilling of the LO metadata with correct information. Metadata is information that describes the characteristics of certain documents, material, or LO. The main purpose of metadata is still to be understood and used by people or software agents in cataloging, searching, and similar tasks (Taylor, 2003). The cataloging and indexation process represents one of the greatest issues to locating educational contents, such as learning objects, because it is through this process that these objects can be found through search engines. Incorrect LO cataloging or indexation causes inefficacy in search processes. This situation is aggravated when LO are distributed and maintained in several distinct repositories. The increase of LO production in Brazil (and around the world) by several different institutions has shown the risk that the material remains unused by the general community, or at least with very restricted use, limited only to the members of the institution in case a unified search mechanism exists capable of finding LO in repositories of most anyone in the institution. Currently there is no standard infrastructure that gives support to a unified search and retrieval of educational resources such as LO (CORDRA Management Group, 2009). To assist in this situation, the present work proposes the creation of an agent-based federated catalog of learning objects (AgCAT). The general objective of this system is to provide an infrastructure of federated LO catalogs that are able to help in the search and retrieval of these educational resources. The system will make intensive use of technologies from Distributed Artificial Intelligence (DAI) and Multi-Agent Systems (MAS) research fields (Weiss, 1999; Wooldridge, 2002), seeking to optimize the LO search process. The system will use several protocols and technologies to harvest metadata from LO repositories and digital libraries. Several AgCAT systems can also be federated, forming a federation of LO catalogs. The search for LO in the federation is transparent for its users. A query made in any federated AgCAT system is transparently propagated to all other AgCAT systems in the federation. Therefore, apart from communication delay, a query in any AgCAT system is equivalent to the same query in any other federated system. Only the search propagation protocol must be supported by each federated AgCAT system. The administration and management of each federated AgCAT system is completely independent from the other federated systems, allowing for different institutions to be included easily in the federation. This work presents the functional structure and organization of the AgCAT system, showing the system's architecture, aspects of its prototype, and main results obtained until now. The next two sections present a literature review concerning the main topics related in the present work focusing on the metadata standards supported by AgCAT and the multi-agent technology that supports the system. The following section describes the multi-agent architecture of the system, the organization of its agents, particular details about the formation of the directory federation, and the metadata harvesting process. …