1. KNOWLEDGE MODELING FOR PROGRAM PLANNING
Presented by Fuhua Lin
Authors: Fuhua Lin Qin Li Dunwei Wen
Alberta, Canada
For ICEB 2005, HONGKONG

2. Outline Introduction Related work Curriculum Petri Nets
Petri Nets Markup Language
Web-based Course Dependency Editor
Conclusions

3. Program Planning for Student Advising
is an interactive decision-making process
assist learners in the development of meaningful learning plans
are compatible with their career/life goals
maximize educational potential through communication and information exchanges with an educator.

4. Adaptive Program Planning
able to assist students to generate a personalized program study plans
possessing knowledge about the domain, program structure, and learners.
One of the important tasks:
Model, represent, and maintain the complex logical relationships among the courses in the program.
There is little effort to formalize the curriculum knowledge representation and manipulation (Zhou, et al., 1996)

5. Petri Nets Based Approach
Petri nets (Petri, 1962) based formal model of program structure knowledge modeling
XML-based representation
Web-based editor for constructing and maintaining curriculum models.

6. Related Work Computer-Aided Student Advising
There have been some standalone expert systems or knowledge-based systems to provide some of the functions of a faculty advisor to students.
Issues
how to acquire and incorporate “real-life” expertise?
how to interface to other systems in the university such as online registration system, academic and student databases?
Most of the student advising systems are research prototypes that focus on only one or two components in the systems, this obscures the need for coordination among the components.

7. Related Work (continued)
Knowledge modeling using Petri nets
validating knowledge bases (Nazareth, 1993) ;
diagnostic knowledge representation and reasoning (Portinale, 1997)
determining requirements (Perdu and Levis, 1993)
representing knowledge in discrete event systems (Muro-Medrano, et al., 1998).
flexible operation planning (Lee & Jung,1994)

8. STM, Graph, OPN, Extended PN
Activity p1 t1 t6 t2 t4 p4 S2 p2 t4 t1 S1 t3 t2 p2 S3 t5 p5
Model
Capability
State Transition Machine
Marked Graph
Ordinary Petri Nets
Curricular Petri Nets
Choices
Yes No
Synchronization
Concurrency
Formalism
Curriculum Knowledge Representation
Low
High

9. Curriculum Petri Nets COMP501 Ready Place
The dynamic behavior of a P/T net is determined by rules concerning the enabling and firing of transitions. When the arcs connecting a transition to its input places have a weight of one, the transition is enabled if all of its input places are marked with at least one token. Only an enabled transition may fire. When it does, one token on each of its input places is removed and one token is created on each of its output places. It is possible, however, for an arc to have a weight greater than one. In this case, if the arc is an input arc then the transition is only enabled if the number of tokens on the place to which it is connected is equal to or greater than its weight. When then transition is fired, the number of tokens removed on the place is equal to the arc’s weight. If it is an output arc, firing the transition creates the number of tokens on the output place equivalent to the arc’s weight.
The numerical values of each of the and functions correspond to the weights of the arcs connecting places and transitions.
By convention arc weights of 1 are not shown explicitly.
Ready Place

10. Concurrency
COMP610 p1
COMP501 p1
COMP667 p1
Done Place

11. Co-requisites-Strict Synchronization
COMP314 p5
COMP200 p1
COMP314_315 p6 p2
COMP315
{m, course 1, course 2, … course n}
If m=-2, the course is a co-requisite of the courses listed
e.g. for course COMP314, another course COMP315 is its co-requisites.
it is value in PNML is -2, COMP315.
In this case, a virtual course is created internally.

12. Choices/Exclusion {-1, course1, course2, …, course n} COMP668 p1
e.g. {-1, COMP697}
COMP696
COMP610 p2
COMP697 p4
COMP695 p3
Choice Place

13. Firing Rules C1 and C2  C3 COMP610 p3 COMP697
Conditional Synchronization!
COMP695 p4
C1 or C2  C3
COMP501 p3
COMP610
COMP601 p4

14. Extending the Standard Firing Rules
To represent curriculum Petri nets using PNML, we do not have any element for the firing rules in the PNML, because we do not use a standard firing rule to fire the course Petri nets automatically. Instead, we set up a special business rule that implemented to fire the transitions of curriculum Petri nets.

15. Boolean Functions Represented Firing Rules
n OUT of m. e.g. {2, C1, C2, C3};
COMP501 p3
COMP689
COMP601 p4
COMP604 p5

16. Firing Rules {n1 OUT of m1} and {n2 OUT of m2} and …
e.g. {1, C1, C2} and {1, C3, C4};
COMP501 p1
COMP601 p2
COMP617
COMP602 p3
COMP603 p4

17. Petri Net Markup Language (PNML)
an XML-based standard format for the interchange of Petri nets.
The main elements in the PNML file for a curriculum are Transition, Place, Arc.

18. PNML-represented Curriculum Model
READYCOMP501
COMP501
<?xml version="1.0" encoding="ISO "?>
<pnml>
<net id="n1" type="BlackTokenNet">
<name>unnamed</name>
<transition id="COMP501"> …
<place id="READYCOMP501">
<arc id="a1" source="READYCOMP501" target="COMP501">
</net>
</pnml>

19. Transition Element in PNML-represented Curriculum Model
READYCOMP501
COMP501
<transition id="COMP501">
<name>
<text> Java Programming </text>
<value>1, READYCOMP501</value>
</name>
</transition>

20. Place Element in PNML-represented Curriculum Model
READYCOMP501
COMP501
<place id=“READYCOMP501">
<marking>
<value>1</value>
</marking>
<name>
<value> READYCOMP501 </value>
</name>
<initialMarking>
</initialMarking>
</place>

21. Arc Element PNML-represented Curriculum Model
READYCOMP501
COMP501
<arc id=“a1“ source =“READYCOMP501” target=“COMP501”>
<inscription>
<direction>input</direction>
<value>1 </value>
</inscription>
</transition>
</arc>

22. <value> in a course transition in PNML…
<value>
m1, course 1, course 2, …, course n1;
m2, course 1, course 2, …, course n2;
ml, course 1, course 2, …, course nl;
</transition>
Notes:
If mi=0, no pre-requisites;
If mi=-1,the current course and the courses listed are mutual-exclusive;
If mi>0, mi courses must have been taken to take the course
If mi=-2, the course is a co-requisite of the courses listed.
In this course a virtual course will be created.

23. Manipulation of the Knowledge
Add(CPN, c)
Delete(CPN, c)
Modify(CPN, c)
Skip(CPN, c)
Select(CPN, c)
Check(CPN)

24. WEB-BASED PROGRAM KNOWLEDGE EDITOR

25. Implementation of the Knowledge Editor
Based on open source package PNK (Java-based Petri Net Kernel)
Save and publish files in PNML (Petri Net Markup Language) for interoperability.
Interacting with users
Java/JSP based HTML Webpage Dialog

26. Application Example: eAdvisor

27. … University 2 University N Learner UI Learner Information Web Service
Notification
Web Service UI
notify
update
Advisor/Registrar
/Administrator
Learner Information
Web Service
request
delegate
Scheduling
Agent DB
Monitoring
Agent
MSc IS Ontology
Web Service
Learner
Data Base UI
monitor
Brokering
Services
delegate/
notify
request/respond
Course Schedule
Web Service
Course Schedule
Web Service
Program/Course
Data Base
Ontology
Course
Schedule
Program/ …
University 2
University N
Program/Course
Data Base

28. Program Structure Represented by Petri Nets
mandatory 1 5
COMP695
p11
COMP636 1
p20 p9
optional m 1
COMP605
p12
p28
p10
COMP617
p21 1 m 1 1
COMP602
p13 1 1
COMP667
p22 1
COMP696 m
COMP603
p14 1 1 1 1 m
COMP648
p23 p1
COMP504 p5 1
COMP697 m
COMP607
p15 1 1 1
p16
COMP660
p24
p29 1 m
COMP610 m p2
COMP503 p6 1
COMP689
p25 1
COMP698 1 m 1 1
p30 p3
COMP501 p7 1 1
p18
COMP641
p26 m
COMP699 m p4
COMP601 p8
COMP604
p17 1
p19
COMP674
p27
p31
Foundation
Core
Career Track
Electives
Graduation
Project/Thesis

29. Interface of eAdvisor

31. Conclusions
For a given program model, the extended Petri nets model is able to
accurately represent the program structure
dynamically reason about “ready-to-take” courses to be used by the program planning agent

32. Future Work
Verification (model-checking) module of curriculum Petri nets based program models
Web Services for Accessing PNML-represented program models.
Modeling concept ontology for developing intelligent e-learning systems.