Tutoring%20

1
Tutoring Help Systems

• Deepthi Bollu for CSE495
• 10/31/2003

2
I Have A Plan Too

• Introduction
• General Aspects of Case Based THS
• Examples
• Perspectives
• ELM-Learner model
• Demo of ELM-ART

3
Take Off

• Tell me and I forget.
• Show me and I remember.
• Involve me and I understand.
• 
  - Chinese
  proverb

4
Introduction

• Traced back to 1926 -Sidney L. Pressey
• TS - provides individualized tutoring or
  instruction.
• Each ITS must have these 3 components
• knowledge of the domain
• knowledge of the learner
• knowledge of teaching strategies

5
Conventional Model Of ITS
6
DRAWBACK

• Limited in its ability to take into account the
  Learners intentions or their Personal problem
  solving Style.
• What can the REMEDIES be then?
• individual info. about how much a particular
  learner solved tasks should be kept for a long
  while.
• This knowledge must be used in subsequent
  diagnoses and tutorial decisions.
• HOW MUCH? HOW LONG?

7
These issues lead to what?

• CASE BASED TUTORING SYSTEM

8

• Where are these cases obtained from?
• Cases produced by the learners themselves.
• Collected from experiences with other learners.
• Cases generated on demand (sub cases that get
  created during problem solving).
• Predefined CasesForeseen valued as helpful by
  experienced tutors.

9
GENERAL ASPECTS OF

• CASE BASED TUTORING SYSTEMS

10
General Aspects Of CBTHS

• Where in THS are CBR techniques used?
• Problem Solving Phase.
• Not just limited to only that but are used in
• Case Based Adaptation.
• Case Based Teaching.

11
General Aspects Of CBTHS

• Case Based Adaptation.
• Adaptable Features for interactive applications.
• So, What are Adaptable systems? Ex CHEF
• Case Based Teaching.
• Main goal-to provide cases that are useful to
  him/her
• to understand new units of knowledge.
• to support Problem Solving.
• Static Case Based Systems, Other systems,
• Adaptive Case based Teaching systems
  (INDIVIDUALISED REMINDINGS).

12
General Aspects Of CBTHS

• Types of Case Based Reasoning Methods
• differ with respect to the goals of these
  Systems
• Classification approach
• (Systems that provide Help )
• Problem Solving approach
• (Systems that support Learning)
• Planning approach
• (Systems that support Planning)

13
Reminder of the topics we learnt

• The Taxonomy of problem solving
• Analysis (interpreting the solution)
• Classification Possible outcomes are known
  in advance.ExYes/No problems
• Diagnosis Not all the outcomes are known in
  advance.Exthe brake lights
• Synthesis( Problem solving / Constructing /
  generating the solution )
• Planning
• Configuration

14
Reminder of the Topics we learnt

• Classes of Adaptation
• Generative solution Adaptation
• Transformational Analogy (ExCHEF)
• Derivational Analogy

15
General Aspects Of CBTHS

• 2 Different Types Of Case Representation (tutorin
  g systems make use of both)
• Complete Cases
• Partial Cases (Snippets) describe Sub goals
  of problems and their Solutions within different
  Contexts. Case is sequence of Planning Decisions.

16
EXAMPLES

• Wide range of applications
• Physics, Biology, Business.Jurisprudence
• Two Areas are of Special Interest.
• CHESS
• PROGRAMMING
• Why are these both areas of special interest?
• No complete Domain theories covering all aspects
  of these domains.

17
PERSPECTIVES

• Showing the examples from the users own learning
  history (how is it useful?)
• On-line Help Systems
• rely not only on introductory questionnaire.
• remembering the last selection user
  made
• but even does this.
• interpret, Store use cases from
  users history

18
Episodic Learner Model
LISP Code
Task Description
Diagnosis (Explanation)
Domain Knowledge
Learner Model
Generalization
Derivation Tree (Explanation Structure)
19

• ELM-learner model used in ELM-PE and ELM-ART that
  stores knowledge about user in terms of
  collection of Episodes (cases)
• Utilized in the diagnostic process
• in the case of complete solutions
• in the case of incomplete solutions
• ELM used
• to analyze solutions
• to retrieve examples and remindings
  to user
• Construction of learner model and architecture of
  ELM.

20
DOMAIN KNOWLEDGE REPRESENTATION

• Thru a hybrid model consisting of CONCEPTS and
  RULES in terms of hierarchically organized
  FRAMES.
• CONCEPTS (instance variables) comprise
• knowledge about programming language
• algorithmic and problem solving knowledge
• CONCEPT FRAMES (classes) contain
• information about plan transformations.(?) rule
  s to solve the goal stated by this concept.
• ADITIONALLY,
• bug rules describing errors.

21
THE DIAGNOSTIC PROCESS

• Starts with TASK DESCRIPTION.
• Concepts will have GOALS can have
  TRANSFORMATIONS.
• ex altering of the order of clauses.
• Every transformation-indexed by set of RULES.
• All rules within CASE are checked and hence
  different EXPLANATIONS.
• DIAGNOSIS recursively called.
• Results in a DERIVATION TREE built from all
  concepts and rules identified to explain the
  learners solution.
• Set of all instances (plus GENERALIZATIONS)
  constitutes the EPISODIC LEARNER MODEL.

22
EXPLANATION BASED RETRIEVAL OF CASES

• Advantage of episodic modeling-Potential to
  predict code that learner will produce
• Predictions used to search for examples and
  remindings useful for solving new task.
• Procedure..
• Generate an expected solution used by Diagnosis
  Component.
• Results in an explanation structure.
• Resulting explanation stored temporarily in the
  ELM.
• Computes similarity to other episodes (examples
  remindings)
• Best match found and offered to the learner as
  example solution
• Temporarily stored case is removed.

23
Putting it all Together
I N T E R F A C E
Learner Model
Comm- unication Model
Domain Know- ledge
Expert Model
Pedagogical Model
24
ELM-PE

• to support novices learning LISP
• consists of
• ALFRED-a Syntax driven Structure Editor
• Intelligent Analysis of task solutions.
• Example based Programming
• Example based Explanation
• limitations of ELM-PE.
• ELM-ART www based LISP course.

25
DEMO of ELM-ART

• http//apsymac33.uni-trier.de8080/elm-art/logi
  n-e

26
CACHET

• CAse based CHess Endgame Tutor

27

• References
• Dr. Munozs lecture classes.
• Various sources over web
• Lecture notes in Case based Reasoning(Weber
  and Schult)
• THANK YOU VERY MUCH