Comprehensive Information Theory Based Intelligent Tutoring System Authoring and Applying Platform
|School||Beijing University of Posts and Telecommunications|
|Course||Signal and Information Processing|
|Keywords||comprehensive information theory intelligent tutoring system authoring tools pattern recognition|
Instruction is one of the most comprehensive information interaction processes among human beings. The education goal oriented communication between instructors and students involves management of pragmatic information of instructing, and the instructors should dynamically adjust the instructing stratage, observing the widespread and complex factors that concerning the learning and instructing state. As to dealing with the complex information processing during instructing, machine intelligence and human intelligence each has its own strengths and weaknesses. This paper’s core purpose is to establish a comprehensive information system supporting the harmonious collaboration of the above two, implemented in the comprehensive authoring and applying platform of intelligent tutoring systems.Since the computer appeared, Continuing and extensive research has been made on intelligent tutoring systems, relative applications gradually entering the community, but there are still some profound barriers. The control information structure is too complex for authoring and maintenance; Compares with human instructors, the machine role make them harder to help students maintain a good learning state. Responsing to these challenges, this work proposed the comprehensive information theory based intelligent tutoring system authoring and applying platform.First of all, a comprehensive information thoery based information structure of intelligent tutoring system was proposed after an in-depth inspecting into the comprehensive information process of education, which included a succinct common-wide information unit design of "status-concept-instant-effectiveness", and employing fuzzy logic solutions to preliminary deal with the concept of uncertainty. Secondly, we proposed a comprehensive information theory based information framework for the self-management and good habits fostering self-support system, a creative integrated solution for raising the learning state of students. Furthermore, by establishing the user-assisted self-time management system, we achieved the coordinated service of various intelligent tutoring systems authored. Then, we presented a set of statistical feature selection based intelligent knowledge mining algorithm, working on the self-time management data and self-aware data, perceiving certain kinds of significance and enlightenment for user, sending further valuable suggestions concerning user’s self-management.As a theoretical basis, we summarize here the concept of comprehensive information. The epistemology level information of something from a subject’s concern, is refers to subject’s perception or expression of something on the movement and its changes, including the form, meaning and effectiveness of its states and the states’ changes. The difference between it and the ontology level information lies in: The epistemology level’s information is from the subject’s concern, with relativity, taking the existence of the subject as the condition; while the latter is the self-expression on the states and the states’ changes of its movement, with the absolute quality and the universality. First, the subject has the sensory capacity, can feel the form of the states and the states’ changes of something’s movement; secondly, the subject has the understanding ability, can understand the inherent meaning of something’s movement; again, the subject has the sense of purpose, thus can judge the value of something’s movement. We called merely the form part as "grammatic information", merely the meaning part as "semantic information", and merely the effectiveness part as "pragmatic information", is called the entire information this kind of rich complete epistemology information’s definition, and we defined such a rich epistemology level information as the comprehensive information.In this paper, the cognitive process is regarded as the process of the subject actively receiving and processing the epistemology level information that is comprehensive information. The subject’s cognitive goal guided the selection of noteworthy dynamic grammmatic information, then this subset was to be dynamic understood as semantic information, in the process of understanding the subject making use of the existing knowledge and reasoning, association, etc. Access to semantic information caused the increase of knowledge, speaking of the cognition goal, the pragmatic information is the extent of the subject’s knowledge accumulation which achieves through this learning process, may be described as the change of the cognitive states of the subject.Instruction is a comprehensive information process, also the the cognitive process of the student as an active subject under the guidance of a teacher. The teacher serves as outside information provider, whose centre for concern is the students’ cognitive activity, with a teaching goal to guide and help students reach their goals of cognition, while the students try to complete the cognitive process making use of the information their teacher provides. As a result of student and teacher’s thought activity, students received all the grammatical information sent by teachers, but the accepted semantic information differed from the understanding of teachers, the actual effectiveness achieved by students should also differ from the planning of teachers, so, we can regard the instruction process as a comprehensive information channel with active effects. What teachers concerned about is to reach the goal of teaching, but how to reach the instruction goal relies on each class’s effect on the student. In order to achieve the goal, the teacher must observe student’s study condition, dynamic alignment teaching process. All the teaching related factors belong to teacher’s concern, such as every lecture’s effect on student’s cognitive process, interest, the environment influence and so on, these information constituted the actual effectiveness set of the instruction contents.On the pedagogical design, the teacher first needs to plan the instructional contents in pragmatic level, according to the student’s cognition effectiveness state and the instructional goal; then, according to the effectiveness requirements, further designs the detailed contents of instruction, such as organizing adequate instructional materials and so on, but it is not until the moment of teaching, can these transform into grammatic information. Generally in human to human teaching, the process of pedagigical design in fact concealed includes these elements, and the teachers had grasped the usage of the instructional modules under various conditions; while in the case of intelligent tutoring system design, the instructional modules were required to contain not only the contents for presentation, but also such control information.In the comprehensive information structure proposed for intelligent tutoring systems, the system is a loose molecular structure as a whole. Each instructional module is similar to an independent member, contained the conditions it suit occurs at, the forecast effectiveness under some representative conditions, as well as the specific instructional materials. They share a mutual state space, on which we has defined the meta-concept set used for expressing various conditions. The specific instructional materials no longer include the control information, which is merely the content presentation. Therefore, so long as the teachers who designed the intelligent tutoring system had publicized the connotations of the conditions and concepts on the mutual state space, other teachers would easily understand its control structure, but also very easy to revise and improve it, for instance to increase the instructional modules, to revise the original condition and effectiveness, and so on.In recent years, this research area started to pay attention to a very significant issue: how to control students’ gaming behavior during the instruction by intelligent tutoring systems, and to improve students’ learning state. The main methods used includes: to give independent suggestion on these behaviors, to refuse certain learner request in particular case, and to improve the software man-machine interaction design quality and so on. This article believed whether a student achieves efficient learning through an intelligent tutoring system, on the one hand relies on the system’s pedagogical design, on the other hand relies on the student’s learning habits. Good learning habits have universal and in-depth influence on a student’s learning, and to form them is a demand of quality education. Intelligent tutoring systems can complement human tutors in supporting or accompanying the process of cultivating good habits by students themselves, for their advantages of computing and precision, time and space superiority, and the role superiority. This article proposed a self-management supporting system, which differs from intelligent tutoring systems, as the former serves as an information mate of a student, understanding the student’s behavior, helping the student to foster good habits, while the latter serve as specialized subject teachers, understood how to teach various disciplines. Systems in the two kinds of role collaborate with each other, and become an harmonious "information companion group" for student’s learning and life.In supporting students to develop good habits, we believe that the formation of a habit relies on self-awareness, at the same time, time management is a necessary foundation. As for achieving the goal to develop good habits, pragmatic information refers to the effectiveness of the student’s experience during a certain period or some certain practice towards habits fostering, and knowledge extraction is needed to obtain it; semantic information refers to the meaning sensed by certain experiences; the information included in the basic self-manage model, such as how the student spend his or her time, all the sort of things in the student’s everyday life, the self-aware records on certain experiences and so on, belongs to grammatical information, which is also the information source for extraction of more complex knowledge at a higher level. The main function of study state supporting system is: let users schedule their learning, thus avoid the condition of learning with temporary motivation and then easily dropping out; be an information center when multiple intelligent tutoring systems serve simultaneously, to help the management of their collaboration; functions for supporting user’s self-management and good habit fostering, for example, monitoring the learning plans on every intelligent tutoring system and give timely tip on adjusting schedule; provide intelligent knowledge extraction services in information companion control center, to further help the user to improve the depth and breadth of self perceptivity, and the user’s self-management level.The integrated system is a comprehensive information theory based intelligent tutoring system authoring and applying platform, which provides bidirectional support: human tutor can use the authoring tool subsystem to design intelligent tutoring systems, while students can use the applying subsystem to use several intelligent tutoring systems coordinated, so as to carry on multi-disciplinary studies, simultaneously carry on self-management and good habit fostering with system’s intelligent support.From the role discrimination, we separately evaluated the two subsystems of the whole platform, which is authoring tools subsystem and applying subsystem. On the one hand, our demand for the authoring tool is to enable teachers to produce rich intelligent tutoring systems more quickly and conveniently, so the experiments and evaluation stressed its development rapidity, succinct and the maintainability; On the other hand, the character of the applying platform lies in supporting collaboration among multiple intelligent tutoring systems and the user’s self-fostering habits, the experiments and evaluation focused on technical aspects, that is the function implementation, quantity of knowledge extracted, the ratio of effective knowledge on the subject’s self-evaluation and so on. An innovative evaluation method was used on the applying platform, which is statistics on performance of intelligent knowledge extraction, with specific population of parameters as premise, for example as took the number of schedule records, everyday matter sorts and self-aware attributes for premise, and did statistics on the important associations discovered by general knowledge extraction, then the ratio of the discovered associations which is new for the user, and the ratio which the user regarded as valuable, the latter was a statistics on users’ subjective view. Results confirmed that it can support rapid development of intelligent tutoring systems without demanding for authors’ programming skill, and the information structured based on comprehensive information theory provided universal support for complex control strategy, the control information was succinct and clear, had good maintainability; the intelligent knowledge extraction part can intelligently discover enlightened and valuable knowledge based on the user’s time-management and self-aware data, by means of intelligent appraisal and suggestion, it can provide users with effective help on self-fostering better learning and life habits.This dissertation first summarized the foundations, such as the comprehensive information theory, the intelligence tutoring system authoring tools and so on; in the second chapter we entered the information science and the technical philosophy perspective, and advanced the research orientation; then followed by a chapter proposed the comprehensive information theory based intelligent tutoring system authoring tool; the fourth chapter addressed extensive discussion on the widely-faced challenge of students’ gaming behavior towards intelligent tutoring systems, and advanced the learning state support system theory; The fifth chapter proposed in-depth information structure of the comprehensive information theory based learning state support system; The sixth chapter discussed the intelligent knowledge extraction on the platform; The seventh chapter carried on the experiments and evaluation analysis, respectively on the intelligent tutoring system authoring aspect and on the intelligent knowledge extraction aspect; in the last chapter, we addressed the work summary and prospects.