Personalized Distance Learning Based on Multiagent Ontological System I. N. Keleberda 1, N. S. Lesna 1, S. D. Makovetskiy 1, V. Terziyan 2 1
Kharkov National University of Radioelectronics, 14, Lenin av., 61166 Kharkov , Ukraine
[email protected] 2 University of Jyvaskyla, P.O. Box35 (Agora), FIN-40014 Jyvaskyla, Finland
[email protected]
Abstract The paper presents architecture of a personalized distance learning system based on multiagent technology and ontological modelling of students’ profiles. Delocalization of a student data in the system is achieved by software agents, which assumed to be distributed at different platforms. These platforms operate as separate Web services and use the ACL (Agent Communication Language) for the data transfer. In this paper the algorithm is proposed, according to which the Multiagent Ontological System for Personalized Distance Learning (MOSPDL) solves the tasks of distant learning process automation, which assume utilization of the ontological models of students’ and learning resources’ profiles.
1: Introduction Recently many distant learning systems have appeared at the market, which provide possibilities to organize more or less successful learning process through the Internet. These systems differ by the scale and the assignment; they have various architectures and are implemented by utilizing different Web technologies. However, the majority of modern distant learning systems are characterized by usage of very restricted set of educational materials. On the other hand, they provide insufficient level of personalization of the learning process and often need physical (PCdependent) localization of a student model data. One possible way for overcoming mentioned difficulties is the usage of multiagent software technologies in the framework of the Semantic Web activities of the W3? consortium [1]. These technologies are capable to extract necessary educational materials (disposed over the whole Web space) to provide high-quality personalization of the distant learning in a PC-independent system.
2: Multiagent ontological system for personalized distance learning (MOSPDL) architecture In this work, a PC-independent architecture of distributed personalized distant learning system (DLS) is proposed. Such architecture gives to a student an access to full set of data collected by the DLS and saved in the student's profile. The architecture is presented in Figure 1. The distinctive feature of the designed architecture is the implementation of delocalization and personalization of the student model data with the help of multiagent ontological approach. Delocalization is provided by software agents [2], which are distributed at different computers. For example, a personal agent is created for each student at the DLS portal; agent-coordinator realizes the handling of the DLS at the server containing current information of learning process; and a learning resources agent provides access to educational materials from computers of the learning service providers. The personalization is achieved at the expense of ontological models, which represent metaknowledge to be used with the DLS. The metaknowledge data are encoding by the specialized markup language, namely OWL (Web Ontology Language) [3]. The model fulfils operations over sets of student's profiles records and educational data distributed in the Internet. The scheme of the proposed MOSPDL algorithm (to handle individual selection of an educational material and operation over the student's profiles with usage of ontological models and educational resources) is as follows. Let PL is the student's profile, and PR is the profile of educational resources. At the intersection of the indicated profiles the choice of an educational material is yielded, which meets the preferences of the student.
which is recommended for the student: A3 (O meta , QLM ) → { li } . Stage 7. At the last stage the agent-coordinator A2 sends the set { li } to the personal agent A1 . The personal agent produces multimedia learning output l for the student: A1 ({ li }) → l . In the next Section, the descriptions of the software agents A1 , A2 , A3 are presented.
3: MOSPDL software agents 3.1: The personal agent
Figure 1. Architecture of the multiagent ontological system for personalized distance learning
The MOSPDL algorithm contains the following stages: Stage 1. The registration of a user in the MOSPDL is provided by means of Web-interface. The personal agent A1 creates the profile of the user: A1 ( I L , O L ) → PL , where I L is the user personal information and OL is the ontological model. The profile PL is written to the database of the DLS. Stage 2. Agent-coordinator A2 sends query QPR for educational data profile PR . Software agent A2 uses metaontology O meta , ontological model O R of the educational resource and the profile PL of the user:
A2 (O meta , OR , PL ) → QPR . Stage 3. The learning resources agent A3 creates the query QM to educational resources in the Internet for searching of educational resources metadata. This software agent uses metaontology O meta and query QPR , namely: A3 (O meta , QPR ) → QM . Stage 4. Educational Internet-resources give metadata for analysis of necessity of their usage in the learning process. The software agent A3 creates the profile PR using information I R and ontological model OR of the educational resources: A3 ( I R , Q R ) → PR . Stage 5. Agent-coordinator A2 provides selection of educational materials from intersection of the profiles PL and PR ; then it sends query Q LM for needed educational materials: A2 (O meta , PR , PL ) → QLM . Stage 6. The learning resources agent A3 on the basis of query QLM build the set of educational materials { li } ,
The main task of a personal agent A1 is creation of a user profile. For this purpose, the scheme of graphical interface creation is used. The user graphical interface is represented in the form of pages, which are dynamically created by the personal agent. These pages come to the student by a Web-browser. Software agent platform works at the DLS as a separate service. Such architecture allows organizing data transfer to the software agents using ACL (Agent Communication Language), appropriate support for which is located in the container of the software agent platform. Algorithmic structure of the software agent A1 contains the stage of registration and the stage of learning. The personal agent realizes a dialogue between the DLS and a student using the graphical interface. However search and primary analysis of selected data will be provided by the learning resources agent, which is described in the next subsection.
3.2: The learning resources agent This software agent belongs to the third link of the chain "Student – DLS – Learning Resources". It plays a role of search engine, which is capable to support search for several resources simultaneously. The task of the learning resources agent A3 consists in searching of educational materials within the scope of a trainee interest. The proposed learning resources agent (as the specialized search engine) is oriented to the direct interaction with other software agents in the MOSPDL. So the pointed agent A3 organizes access to the database containing educational materials. An interaction of the learning resources agent A3 with a database is organized by means of the JDBC-driver of the selected database management system. This provides a minimal dependency of the designed MOSPDL on a database management system.
Algorithmic structure of agent A3 contains the stage of forming the profiles for educational materials and the stage of creation of the needed educational materials set. The stage of forming of the profiles for educational materials consists of three steps: 1. Accessing the information on educational materials at a learning resource and structuring the requested information on the basis of the IEEE 1484.12.1 Standard for Information Technology of Education and Training Systems – Learning Objects and Metadata (LOM) [4]. 2. Creating the set of the profiles of educational material, which contain the metadata about appropriate educational material. 3. Transmitting the generated profiles of educational materials to the agent-coordinator. The stage of creation of the needed educational materials set consists of two steps: 1. Creation of the set of educational materials according to a query. 2. Transmitting the indicated set of educational materials to the agent-coordinator by means of the interaction protocol. The interaction between personal agent A1 and learning resources agent A3 , individual selection of an educational material and, finally, the control over the learning process is realized by software agent-coordinator A2 .
2. The templates are filled on the basis of a resource model and the student's profile. The queries are aimed to get the educational materials, which contain metadata in accordance with the IEEE 1484.12.1 Standard for Learning Object Metadata (LOM) [4]. 3. The message to the learning resources agent is formed, which includes query regarding the profiles of the educational materials. The stage of individual selection of an educational material consists of the following steps: 1. In accordance with returned profiles of educational materials, the agent-coordinator fulfils comparison of these profiles with the student's profile. It allows generating a set of the references on a required educational material oriented to the individual characteristics of the student. 2. The query to the learning resources agent is created regarding necessary educational material set based on the indicated set of the references. 3. The query message to the learning resources agent (regarding necessary educational material set) is transmitted. 4. The downloading of necessary educational material set is fulfilled. The main function of the agent-coordinator is to personalize the distant learning process by means of individual selection of educational material.
4: Conclusions 3.3: The agent-coordinator The agent-coordinator provides functions of a mediator and realizes control over the learning process in the MOSPDL. Collection of the personal agents, the learning resources agents and the agent-coordinator represents the whole multiagent system. The interaction of the software agents is carried out on the basis of the transfer protocol according to the ACL specification. This protocol supports XML (eXtensible Markup Language) format. The software agents use ontological models to support both forming of queries and processing of the retrieved data. This allows working with semantics of the indicated queries at the level of the transmitted information concepts. Algorithmic structure of the agent-coordinator A2 contains the stage of searching for educational materials and the stage of individual selection of an educational material. The stage of searching for educational materials consists of the following steps: 1. The queries are created (with usage of appropriate templates) and delivered to the set of different agents of the educational resources. Types of the templates are represented in the metaontology.
The approach and designed agent-based software system suppose to belong to a new generation of distributed systems of distant Web-based learning, namely to Semantic Web-enabled multiagent ontological systems. The elaborated architecture and algorithm of MOSPDL is intended to solve the task of automation of the distant learning process, which is oriented on utilizing ontological models of student's profiles and learned resources profiles.
References [1] T. Berners-Lee, J. Hendler, O. Lassila, "The Semantic Web", Scientific American, Vol. 284, No. 5, 2001, pp.34–43. [2] D.A. Milashenko, S.D. Makovetskiy, R.V. Boblovskiy, I.N. Keleberda, N.S. Lesna, "Software Agents for Learning Resources of Digital Library", Lecture Notes in Informatics, Vol.P30, Gesselschaft für Informatik, Bonn, 2003, pp.77–84. [3] OWL Web Ontology Language Reference, Available at http://www.w3.org/TR/owl-ref/ [4] IEEE 1484.12.1 Standard for Learning Object Metadata (Standard for Information Technology of Education and Training Systems – Learning Objects and Metadata), 2002.