A software based mentor system

Abstract

This thesis describes the architecture, implementation issues and evaluation of Mentor - an educational support system designed to mentor students in their university studies. Students can ask (by typing) natural language questions and Mentor will use several educational paradigms to present information from its Knowledge Base or from data-mined online Web sites to respond. Typically the questions focus on the student’s assignments or in their preparation for their examinations. Mentor is also pro-active in that it prompts the student with questions such as "Have you started your assignment yet?". If the student responds and enters into a dialogue with Mentor, then, based upon the student’s questions and answers, it guides them through a Directed Learning Path planned by the lecturer, specific to that assessment. The objectives of the research were to determine if such a system could be designed, developed and applied in a large-scale, real-world environment and to determine if the resulting system was beneficial to students using it. The study was significant in that it provided an analysis of the design and implementation of the system as well as a detailed evaluation of its use. This research integrated the Computer Science disciplines of network communication, natural language parsing, user interface design and software agents, together with pedagogies from the Computer Aided Instruction and Intelligent Tutoring System fields of Education. Collectively, these disciplines provide the foundation for the two main thesis research areas of Dialogue Management and Tutorial Dialogue Systems. The development and analysis of the Mentor System required the design and implementation of an easy to use text based interface as well as a hyper- and multi-media graphical user interface, a client-server system, and a dialogue management system based on an extensible kernel. The multi-user Java-based client-server system used Perl-5 Regular Expression pattern matching for Natural Language Parsing along with a state-based Dialogue Manager and a Knowledge Base marked up using the XML-based Virtual Human Markup Language. The kernel was also used in other Dialogue Management applications such as with computer generated Talking Heads. The system also enabled a user to easily program their own knowledge into the Knowledge Base as well as to program new information retrieval or management tasks so that the system could grow with the user. The overall framework to integrate and manage the above components into a usable system employed suitable educational pedagogies that helped in the student’s learning process. The thesis outlines the learning paradigms used in, and summarises the evaluation of, three course-based Case Studies of university students’ perception of the system to see how effective and useful it was, and whether students benefited from using it. This thesis will demonstrate that Mentor met its objectives and was very successful in helping students with their university studies. As one participant indicated: ‘I couldn’t have done without it.’