Disciple Cog Demo

Embed Size (px)

Citation preview

  • 7/28/2019 Disciple Cog Demo

    1/99

    PRESENTED BY

    LEARNING AGENTS

    LABORATORY

    GEORGE MASON UNIVERSITY

  • 7/28/2019 Disciple Cog Demo

    2/99

    Disciple-RKF/COG:

    Agent Teaching by Subject Matter Experts

  • 7/28/2019 Disciple Cog Demo

    3/99

    SPONSORED BY

  • 7/28/2019 Disciple Cog Demo

    4/99

    The Summary of the DEMO

    About Disciple-COG

    Students use of Disciple-COG as a learning assistant

    1. Specifying a scenario2. Viewing the solutions generated by Disciple

    3. Generating a center of gravity analysis report

    Teaching Disciple-COG by a subject matter expert1. Specifying a training scenario

    2. Modeling experts reasoning

    3. Task and rule learning

    4. Rule refinement

    5. Exception handling6. Problem solving

    Tools for the knowledge engineer

    1. Ontology development tools

    2. Ontology import tools

  • 7/28/2019 Disciple Cog Demo

    5/99

    About Disciple-COG

    Disciple is a theory, methodology, and learning

    agent shell for rapid development of knowledgebases and agents, by subject matter experts, with

    limited assistance from knowledge engineers

    Disciple learns from the expert,

    building, verifying and improving

    its knowledge base

    Interf

    ace

    The expert teaches Disciple in

    a way that resembles how theexpert would teach a person. DISCIPLE SYSTEM

    Problem

    Solving

    Ontology+ RulesLearning

  • 7/28/2019 Disciple Cog Demo

    6/99

    Demo overview

    First we will demonstrate how a developed Disciple

    agent helps the students at the US Army War College to

    learn about Center of Gravity analysis.

    Then, in the main part of the demo, we will show how

    this Disciple agent was developed by being taught the

    problem solving method of a military expert.

    Finally, we will demonstrate additional tools that are

    used by the knowledge engineer to perform knowledge

    base development tasks that are currently beyond the

    capabilities of a subject matter expert.

  • 7/28/2019 Disciple Cog Demo

    7/99

    Students use of Disciple-COG

    as a learning assistant

    In the Case Studies in Center of Gravity Analysis

    course students learn to identify the centers of gravity

    (COG) of the opposing forces in military conflicts.

    The COG of a force is its main source of strength,

    power, and resistance.

    Each student has to study a historic scenario (such as

    the World War II invasion of the island of Sicily by theAllied Forces) and has to develop a center of gravity

    analysis report.

    Disciple guides the students to specify the relevantaspects of the assigned war scenario. Then it identifies

    and tests the strategic center of gravity candidates for

    that scenario, and generates a center of gravity analysis

    report. The students study and critique the solutions

    generated by Disciple and finalize the report.

  • 7/28/2019 Disciple Cog Demo

    8/99

    1. Scenario Elicitation

    First the student is guided by thescenario elicitation tool to specify

    the relevant aspects of the war

    scenario.

  • 7/28/2019 Disciple Cog Demo

    9/99

    1. Scenario Elicitation

    The tool shows:

    A table of contents that

    will be developed during

    scenario elicitation.

    A list of

    questions that

    have to be

    answered by the

    student.

  • 7/28/2019 Disciple Cog Demo

    10/99

    1. Scenario Elicitation

    The student is asked to

    provide a short Englishdescription of the

    scenario.

  • 7/28/2019 Disciple Cog Demo

    11/99

    1. Scenario Elicitation

    The student is asked to

    provide a short Englishdescription of the

    scenario.

  • 7/28/2019 Disciple Cog Demo

    12/99

    1. Scenario Elicitation

    Disciple assists by

    providing clarificationsfor each requested

    piece of information.

    Then the student has to

    name the opposing

    forces of the scenario.

  • 7/28/2019 Disciple Cog Demo

    13/99

    1. Scenario Elicitation

    Each opposing force is

    automatically

    introduced into the

    table of contents

    When the student

    selects one of these

    forces, Disciple

    asks more specificquestions about it.

  • 7/28/2019 Disciple Cog Demo

    14/99

    1. Scenario Elicitation

    Because the student has

    characterized Allied_Forces_1943

    as a multi state force, Discipleexpands the table of contents

    with the aspects that are relevant

    for such a force.

  • 7/28/2019 Disciple Cog Demo

    15/99

    1. Scenario Elicitation

    When the student selects one ofthese aspects, Disciple asks

    more specific questions about

    it.

  • 7/28/2019 Disciple Cog Demo

    16/99

    1. Scenario Elicitation

    In this way the student is guided to

    specify the relevant aspects of the

    scenario.

  • 7/28/2019 Disciple Cog Demo

    17/99

    2. Viewing the solutions generated by Disciple

    After the scenario is specified

    Disciple can identify and test the

    corresponding strategic center of

    gravity candidates

    2 Vi i th l ti t d b Di i l

  • 7/28/2019 Disciple Cog Demo

    18/99

    2. Viewing the solutions generated by Disciple

    The tool shows:

    The strategic center of

    gravity candidates for

    each of the opposing

    forcesA description of

    the selectedcandidate and

    the rationale for

    its identification

    and testing

    2 Vi i th l ti t d b Di i l

  • 7/28/2019 Disciple Cog Demo

    19/99

    2. Viewing the solutions generated by Disciple

    The solution includes a

    summary of the specific

    candidates solution.

    2 Vi i th l ti t d b Di i l

  • 7/28/2019 Disciple Cog Demo

    20/99

    2. Viewing the solutions generated by Disciple

    A justification for the entities

    selection as a candidate.

    2 Viewing the solutions generated by Disciple

  • 7/28/2019 Disciple Cog Demo

    21/99

    2. Viewing the solutions generated by Disciple

    And a summary of the

    candidates testing.

    2 Viewing the solutions generated by Disciple

  • 7/28/2019 Disciple Cog Demo

    22/99

    2. Viewing the solutions generated by Disciple

    The

    justificationmay be

    presented in a

    more abstract

    form

    2 Viewing the solutions generated by Disciple

  • 7/28/2019 Disciple Cog Demo

    23/99

    2. Viewing the solutions generated by Disciple

    Or it may be

    presented in a

    more detailed form

    2 Viewing the solutions generated by Disciple

  • 7/28/2019 Disciple Cog Demo

    24/99

    2. Viewing the solutions generated by Disciple

    Or it may be

    presented in a

    more detailed form

    3 Generating a center of gravity analysis report

  • 7/28/2019 Disciple Cog Demo

    25/99

    3. Generating a center of gravity analysis report

    Then Disciple generates a draftcenter of gravity analysis report.

    3 Generating a center of gravity analysis report

  • 7/28/2019 Disciple Cog Demo

    26/99

    3. Generating a center of gravity analysis report

    The first part of the report is a description of the scenario

    which is based on the students specification.The student may further improve it by using a text editor.

  • 7/28/2019 Disciple Cog Demo

    27/99

  • 7/28/2019 Disciple Cog Demo

    28/99

  • 7/28/2019 Disciple Cog Demo

    29/99

    1. Specifying a training scenario

  • 7/28/2019 Disciple Cog Demo

    30/99

    p y g g

    Based on the information elicited from

    the expert Disciple creates a formalrepresentation of the scenario.

    The internal representation

    may be viewed usingthe Association Browser

    2. Modeling experts reasoning

  • 7/28/2019 Disciple Cog Demo

    31/99

    g p g

    Then the expert has to express his reasoning in center of gravity

    identification and testing for the Okinawa scenario,

    by using the task reduction paradigm.

    Identify and test a strategic COG candidate for the Okinawa_1945 scenario

    Okinawa_1945 is a major theater of war scenario

    Identify and test a strategic COG candidate for Okinawa_1945 which is a major theater of war scenario

    What kind of scenario is Okinawa_1945?

    I need to

    Therefore I need to

    Which is an opposing force in the Okinawa_1945 scenario?

    Japan_1945

    Identify and test a strategic COG candidate for Japan_1945

    Therefore I need to

  • 7/28/2019 Disciple Cog Demo

    32/99

    2. Modeling experts reasoning

  • 7/28/2019 Disciple Cog Demo

    33/99

    To reduce the current

    task the expert needs

    some information that is

    identified by asking aquestion.

    The first steps of the reasoning process show how the expert has identified

    Emperor Hirohito as a strategic center of gravity candidate for Japan.We will now show how the expert uses the Modeling tool to express his

    reasoning of testing whether Emperor Hirohito is a viable candidate.

    2. Modeling experts reasoning

  • 7/28/2019 Disciple Cog Demo

    34/99

    To reduce the current

    task the expert needs

    some information that is

    identified by asking aquestion.

    Then the expert must

    provide an answer to

    this question

    2. Modeling experts reasoning

  • 7/28/2019 Disciple Cog Demo

    35/99

    Then the expert must

    provide an answer to

    this question

    2. Modeling experts reasoning

  • 7/28/2019 Disciple Cog Demo

    36/99

    Then the expert must

    provide an answer to

    this question

    2. Modeling experts reasoning

  • 7/28/2019 Disciple Cog Demo

    37/99

  • 7/28/2019 Disciple Cog Demo

    38/99

    2. Modeling experts reasoning

  • 7/28/2019 Disciple Cog Demo

    39/99

    The question and its

    answer lead to this

    reduction of the top task.

    2. Modeling experts reasoning

  • 7/28/2019 Disciple Cog Demo

    40/99

  • 7/28/2019 Disciple Cog Demo

    41/99

    2. Modeling experts reasoning

  • 7/28/2019 Disciple Cog Demo

    42/99

    2. Modeling experts reasoning

  • 7/28/2019 Disciple Cog Demo

    43/99

    2. Modeling experts reasoning

  • 7/28/2019 Disciple Cog Demo

    44/99

  • 7/28/2019 Disciple Cog Demo

    45/99

  • 7/28/2019 Disciple Cog Demo

    46/99

    2. Modeling experts reasoning

  • 7/28/2019 Disciple Cog Demo

    47/99

    This process continues in the same way until the expert has enough

    information to either eliminate or not eliminate the COG candidate.

  • 7/28/2019 Disciple Cog Demo

    48/99

  • 7/28/2019 Disciple Cog Demo

    49/99

    3. Task and Rule Learning

  • 7/28/2019 Disciple Cog Demo

    50/99

    First the expert and Disciple have to formalize theEnglish statements of the tasks.

    This is done in the

    Formalization mode.

    3. Task and Rule Learning

  • 7/28/2019 Disciple Cog Demo

    51/99

    In the formalization mode the tool shows:

    The modeling

    in English

    The formalized

    tasks

  • 7/28/2019 Disciple Cog Demo

    52/99

    3. Task and Rule Learning

  • 7/28/2019 Disciple Cog Demo

    53/99

    Disciple will propose a

    formalization of the task

    The expert may

    accept it or

    he may edit it

  • 7/28/2019 Disciple Cog Demo

    54/99

  • 7/28/2019 Disciple Cog Demo

    55/99

  • 7/28/2019 Disciple Cog Demo

    56/99

  • 7/28/2019 Disciple Cog Demo

    57/99

  • 7/28/2019 Disciple Cog Demo

    58/99

  • 7/28/2019 Disciple Cog Demo

    59/99

  • 7/28/2019 Disciple Cog Demo

    60/99

  • 7/28/2019 Disciple Cog Demo

    61/99

    3. Task and Rule Learning

  • 7/28/2019 Disciple Cog Demo

    62/99

    The informal structure is

    used in the communication

    with the user.

    This is the general

    task reduction

    rule learned by

    Disciple.

    In addition to the

    formal structure of

    the rule, which is

    used in problem

    solving and learning,

    Disciple maintainsalso an informal

    structure of the rule.

  • 7/28/2019 Disciple Cog Demo

    63/99

  • 7/28/2019 Disciple Cog Demo

    64/99

  • 7/28/2019 Disciple Cog Demo

    65/99

  • 7/28/2019 Disciple Cog Demo

    66/99

  • 7/28/2019 Disciple Cog Demo

    67/99

  • 7/28/2019 Disciple Cog Demo

    68/99

  • 7/28/2019 Disciple Cog Demo

    69/99

  • 7/28/2019 Disciple Cog Demo

    70/99

  • 7/28/2019 Disciple Cog Demo

    71/99

  • 7/28/2019 Disciple Cog Demo

    72/99

  • 7/28/2019 Disciple Cog Demo

    73/99

    4. Rule Refinement

    The expert has to examine this step

  • 7/28/2019 Disciple Cog Demo

    74/99

    correct and completely

    explainedby selecting Correct Example

    correct but incompletely

    explainedby selecting Explain Example

    incorrectby selecting Incorrect Example

    p p

    and has to indicate whether it is:

  • 7/28/2019 Disciple Cog Demo

    75/99

  • 7/28/2019 Disciple Cog Demo

    76/99

  • 7/28/2019 Disciple Cog Demo

    77/99

  • 7/28/2019 Disciple Cog Demo

    78/99

  • 7/28/2019 Disciple Cog Demo

    79/99

  • 7/28/2019 Disciple Cog Demo

    80/99

    Disciple guides the

    expert in the definition

    of new knowledge about

    5. Exception Handling

  • 7/28/2019 Disciple Cog Demo

    81/99

    g

    the selected candidate

  • 7/28/2019 Disciple Cog Demo

    82/99

  • 7/28/2019 Disciple Cog Demo

    83/99

    5. Exception Handling

  • 7/28/2019 Disciple Cog Demo

    84/99

  • 7/28/2019 Disciple Cog Demo

    85/99

  • 7/28/2019 Disciple Cog Demo

    86/99

  • 7/28/2019 Disciple Cog Demo

    87/99

  • 7/28/2019 Disciple Cog Demo

    88/99

  • 7/28/2019 Disciple Cog Demo

    89/99

  • 7/28/2019 Disciple Cog Demo

    90/99

    2. Ontology import tools

  • 7/28/2019 Disciple Cog Demo

    91/99

    Ontology import tools perform a three step import process:

    Mixed-initiative retrieval of potentially relevant ontological

    knowledge from external knowledge repositories, such as CYC;

    Automatic translation of the retrieved ontological knowledge into

    an intermediate Disciple ontology;

    Mixed-initiative import from the intermediate Disciple ontologyinto the destination Disciple ontology.

    The following is a brief demonstration of ontology import from

    CYC.

  • 7/28/2019 Disciple Cog Demo

    92/99

  • 7/28/2019 Disciple Cog Demo

    93/99

    2. Ontology import tools

    The knowledge retrieved from CYC is automatically translated into

    an intermediate Disciple ontology

  • 7/28/2019 Disciple Cog Demo

    94/99

    an intermediate Disciple ontology.

  • 7/28/2019 Disciple Cog Demo

    95/99

  • 7/28/2019 Disciple Cog Demo

    96/99

    2. Ontology import tools

    The ontology fragment represented by NATO, its super-concepts,

    and its features is copied into the destination Disciple ontology

  • 7/28/2019 Disciple Cog Demo

    97/99

    A + denotes a term

    that does not yet

    exist in the ontology,while a + denotes an

    existing term

    The user specifiesthe position of the

    imported concepts in

    the destination

    Disciple ontology

  • 7/28/2019 Disciple Cog Demo

    98/99

  • 7/28/2019 Disciple Cog Demo

    99/99