Pitt Logo LinkContact Us

Using Natural Language as Knowledge Representation in an Intelligent Tutoring System

Jung, Sung-Young (2012) Using Natural Language as Knowledge Representation in an Intelligent Tutoring System. Doctoral Dissertation, University of Pittsburgh.

[img]
Preview
PDF - Primary Text
Download (3029Kb) | Preview

    Abstract

    Knowledge used in an intelligent tutoring system to teach students is usually acquired from authors who are experts in the domain. A problem is that they cannot directly add and update knowledge if they don’t learn formal language used in the system. Using natural language to represent knowledge can allow authors to update knowledge easily. This thesis presents a new approach to use unconstrained natural language as knowledge representation for a physics tutoring system so that non-programmers can add knowledge without learning a new knowledge representation. This approach allows domain experts to add not only problem statements, but also background knowledge such as commonsense and domain knowledge including principles in natural language. Rather than translating into a formal language, natural language representation is directly used in inference so that domain experts can understand the internal process, detect knowledge bugs, and revise the knowledgebase easily. In authoring task studies with the new system based on this approach, it was shown that the size of added knowledge was small enough for a domain expert to add, and converged to near zero as more problems were added in one mental model test. After entering the no-new-knowledge state in the test, 5 out of 13 problems (38 percent) were automatically solved by the system without adding new knowledge.


    Share

    Citation/Export:
    Social Networking:

    Details

    Item Type: University of Pittsburgh ETD
    ETD Committee:
    ETD Committee TypeCommittee MemberEmailORCID
    Committee ChairLesgold, Alanal@pitt.edu
    Committee MemberVanLehn, KurtKurt.Vanlehn@asu.edu
    Committee MemberAshley, Kevinashley@pitt.edu
    Committee MemberHe, Daqinggaorenhepitt@126.com
    Title: Using Natural Language as Knowledge Representation in an Intelligent Tutoring System
    Status: Published
    Abstract: Knowledge used in an intelligent tutoring system to teach students is usually acquired from authors who are experts in the domain. A problem is that they cannot directly add and update knowledge if they don’t learn formal language used in the system. Using natural language to represent knowledge can allow authors to update knowledge easily. This thesis presents a new approach to use unconstrained natural language as knowledge representation for a physics tutoring system so that non-programmers can add knowledge without learning a new knowledge representation. This approach allows domain experts to add not only problem statements, but also background knowledge such as commonsense and domain knowledge including principles in natural language. Rather than translating into a formal language, natural language representation is directly used in inference so that domain experts can understand the internal process, detect knowledge bugs, and revise the knowledgebase easily. In authoring task studies with the new system based on this approach, it was shown that the size of added knowledge was small enough for a domain expert to add, and converged to near zero as more problems were added in one mental model test. After entering the no-new-knowledge state in the test, 5 out of 13 problems (38 percent) were automatically solved by the system without adding new knowledge.
    Date: 01 February 2012
    Date Type: Publication
    Defense Date: 29 August 2011
    Approval Date: 01 February 2012
    Submission Date: 10 November 2011
    Release Date: 01 February 2012
    Access Restriction: No restriction; The work is available for access worldwide immediately.
    Patent pending: No
    Number of Pages: 101
    Institution: University of Pittsburgh
    Thesis Type: Doctoral Dissertation
    Refereed: Yes
    Related URLs:
    Degree: PhD - Doctor of Philosophy
    Uncontrolled Keywords: knowledge representation, knowledge acquisition, natural language understanding, intelligent tutoring system. authoring system, physics problem solving
    Schools and Programs: Dietrich School of Arts and Sciences > Intelligent Systems
    Date Deposited: 01 Feb 2012 07:30
    Last Modified: 16 Jul 2014 17:02

    Actions (login required)

    View Item

    Document Downloads