Link System in Concept Map

1
Link System in Concept Map

• Research proposal for GetSmart Project
• Yiwen Zhang
• August, 2002

2
Agenda

• Introduction
• Literature review
• Research questions
• Proposed approaches
• Testbed and experiments design
• Timeline

3

• Introduction
• Literature review
• Research questions
• Proposed approaches
• Testbed and experiments design
• Timeline

4
Concept map

• Concept map is a graphical representation of
  knowledge.
• Nodes (points or vertices) represent concepts
• Links (arcs or lines) represent the relationships
  between concepts
• In the 1960s, Joseph D. Novak at Cornell
  University began to study concept mapping
  techniques.
• His work was based on David Ausubels theories
  (1968) --meaningful learning. Meaningful learning
  involves the assimilation of new concepts and
  propositions into existing cognitive structures.

5
An example
6
Research

• To aid individual learning
• Joseph D. Novak, Apartment of Education, Cornell
  University ( also research scientist at Institute
  of Human Machine Cognition, University of West
  Florida
• Learning How To Learn, New York Cambridge
  University Press. 1984.
• The Theory Underlying Concept Maps and How To
  Construct Them, on web, 2001
• To aid communication of complex ideas in groups
• William M.K. Trochim, Professor in the Department
  of Policy Analysis and Management at Cornell
  University
• Concept Mapping for Evaluation and Planning.
  Evaluation and Program Planning v12 n1 spec
  issue p1-111 1989
• To aid instruction and evaluation
• National Center on Research on Evaluation,
  Standards and Student Testing

7
New trends in research

• Background
• Widespread of computers and internet
• Increasing interest on the research of digital
  library and knowledge engineering
• New research
• Use concept maps to organize hypermedia resources
• Use concept maps for knowledge construction,
  sharing and, intelligent information retrieval
• Major researchers
• Cañas, A. J., K. M. Ford, Institute of Human
  Machine Cognition, University of West Florida
  (CMap)
• Brian R Gaines and Mildred L G Shaw, Knowledge
  Science Institute, University of Calgary (Kmap)
• Subtopics
• Online collaboration
• Expert concept maps and knowledge modeling
• Open learning environment and ontology
  construction
• The process of naming links

8

• Introduction
• Literature review
• Research questions
• Proposed approaches
• Testbed and experiments design
• Timeline

9
Link systems in concept maps

• Link names indicate the users understanding of
  the relationships between concepts
• Two approaches in naming links in concept maps
• Open link system link names are generated by
  users
• Closed link system users select names from a
  given list of link names
• Current research
• Open systems dominate

1. The influence of learner differences on the
construction of hypermedia concepts a case
study. J.M. Oughton, W.M. Reed, Computers in
Human Behavior 15(1999) 11-50
10
Comparison of the two approaches

• Open link systems
• Allow free thinking and expression
• Closed link systems
• For map creators Help to clarify relationships
  between pairs of concepts
• Require critical thinking
• For map readers Help to share and collaborate
• However, little papers discussed the design and
  usage of link systems
• There is no common agreement on a certain set of
  link systems
• There is a concern about whether controlled link
  will limit the flexibility in building concept
  maps, thus hampering the learning process

11
Comparison of the two approaches

• The open link systems have been major obstacles
  for several research topics in concept maps,
  since they
• Involve a lot of manual work in evaluation
• Cause confusion in map sharing
• Have no reasoning function in knowledge
  construction
• There has been a growing consensus among
  researchers that links should be named,
  modifiable, directional, and represented by
  canonical sets.
• --Reliability and validity of a computer-based
  knowledge mapping system to measure content
  understanding (H. E. Herl, H. F. ONeil Jr.,
  Computers in Human Behavior 15 (1999) )

12
Canonical link types link names

• Collins and Quillian a system of relational
  categories (1972)
• Superset is a , is a member of
• Subset consists of, contains
• Similarity is like, is not like
• Part part of
• Proximity is adjacent to, is next to
• Consequence leads to, influence, cause
• Precedence prior to
• Comments
• Widely cited on the research related to semantic
  network
• Proximity is vague comparing to others

Collins, A.M., Quillian, M.R. (1972). How to
make a language user. In E. Tulving, W.
Donaldson (Eds.), Organization of memory (pp.
309-351). New York Academic Press
13
Canonical link types link names (cont.)

• Dansereau and Holley a 3-category link system
  (1982)
• Chaining leads to, results in, produces
• Clustering is like, property of, evidence of
• Hierarchical part of, example of
• Comments
• A higher-level categorization
• Not very clear, since some link names under one
  category are quite different

Dansereau, D. F., Collins, C. D.(1982),
Development and evaluation of a text mapping
strategy. In A. Flammer,, W. Kintsch (Eds.),
Discourse processing. Amsterdam North Holland
Publishing
14
Canonical link types link names (cont.)

• Lambiotte (1989)
• Hierarchy
• Chain
• Cluster
• Procedural
• Influence
• Part
• Comments
• Cluster and chain are appropriate for the
  domains of chemistry and biology to which concept
  maps were first introduced, while they are not
  appropriate in other domains

Lambiotte, J.G., Dansereau, D.F., Cross, D.R.
Reynolds, S.B. (1989). Multi-relational semantic
maps. Educational Psychology Review, 1(4),
331-367.
15
Link systems in education

• Harmon, S.W., Dinsmore, S. (1994)
• Exemplary (knowledge level)
• Associative (comprehensive level)
• Similar Comparative (application level)
• Opposite comparative (Synthesis level)
• Componential (Analysis level)
• Causal ( Evaluation level)
• Sequential
• Comments
• Reasonable and clear in general
• Associative is too vague
• Using exemplary instead of hierarchy can help
  to understand a abstract concept while missing
  some other relationships
• The later work of Oughton Reed (1999) to match
  the link and level of cognitive learning in
  Blooms taxonomy has not been justified, but was
  interesting.

Harmon, S.W., Dinsmore, S. (1994). Novice
Linking in Hypermedia Environments. Paper
presented at the 34th International Conference of
the Association for the Development of
Computer-based Instructional Systems, Norfolk,
VA.. 10(1-4).
J.M. Oughton, W.M. Reed, The influence of learner
differences on the construction of hypermedia
concepts a case study. Computers in Human
Behavior 15(1999) 11-50
16
Link systems in education (cont.)
Terence R. Smith (2002)

• A framework to describe some special information
  needed in the domain of mathematics, science and
  engineering (MSE)
• An incompleted list selected from that framework
  
• Knowledge Domain
• Historical Origins
• Definition
• Scientific use
• Representation
• Defining operation
• Property
• Comments
• The whole framework is not well organized
• The list is valuable to facilitate learning
• How can we integrate them into a link system?

Terence R. Smith, Structured Models of Scientific
Concepts for Organizing, Accessing, and using
Learning Materials, JCDL, 2002
17
Semantic Network in AI

• A graphic notation for representing knowledge in
  patterns of interconnected nodes and arcs, that
  can
• represent knowledge
• support automated systems for reasoning about
  knowledge.
• History
• have long been used in philosophy, psychology,
  and linguistics
• Quillian (1968)
• Woods (1975), "What's in a Link Foundations for
  Semantic Networks" Brachman (1977), Whats in a
  concept structural foundations for semantic
  nets.
• Variety
• a family of representational schemes rather than
  a single formalism
• Example
• WordNet Synonymy, Antonymy, Hyponymy, Meronymy,
  Troponomy, Entailment
• KL-ONE, CLASSIC

18
Classification and ontology in Library science

• From a KR perspective, library scientists are in
  several ways the perfect collaborators. and the
  first practicing ontologist, as evidenced by the
  dewey decimal system and other similar
  classification systems.
• Classification (e.g.)
• , 400 Language
• 401 Philosophy theory
• 402 Miscellany
• 403 Dictionaries encyclopedias
• 404 Special topics
• 405 Serial publications,
• Ontology (e.g.)
• Author, publisher, date,

ScholOnto An Ontology-Based Digital Library
Server for Research Documents and Discourse,
International. Journal on Digital Libraries, 3
(3) (August/Sept., 2000), Springer-Verlag
19
Comparisons and Implications

• Common points
• Knowledge representation
• Composed of concepts and relationships
• Be used interactively

• Implications
• Can concept map be used for reasoning?
• Can the process of creating concept maps also be
  a process of building Digital Library?

20
Summary and research motivations

• In the approaches that use closed link systems,
  we found that
• Some link types are commonly agreed
• Any single set of link types and names cant fit
  into every domain.
• Implications
• Its possible to design a set of link types, but
  impossible and unnecessary to define link names
• We need to consider some domain-specific link
  names
• Can we identify or design a link system that
• Can aid user in learning at least as effective as
  open link systems
• Can partly solve the limitation of open link
  systems

21

• Introduction
• Literature review
• Research questions
• Proposed approaches
• Testbed and experiments design
• Timeline

22
Research questions

• Question1
• Can we propose a link system that can be used in
  concept map construction to assist users
  learning more effectively than open link systems?
• Question 2
• Can the above link system assist instructors in
  instruction more effectively?
• Question 3
• Can the above link system support reuse of
  knowledge repository organized by concept maps
  more effectively and efficiently?

23

• Introduction
• Literature review
• Research questions
• Proposed approaches
• Testbed and experiment design
• Timeline

24
Proposed approaches

• A semi-open link system
• 2-Level
• 1st level predefined set of link types
• All the link names can be categorized into these
  types
• 2nd level suggested link names
• Identify several link names under each link type

25
Link system design

• Step1 Find the link types under canonical link
  types
• Step2 Suggest several link names for each type
• Step3 Consider some domain specific link names
• Put it under the related link type
• Build new link types character, function and
  evaluation

26
Link system design
Character defining operation has operation
property has (color, weight, etc) Function
scientific use applied in, express,
use Evaluation has cons, has pros
27
An example
hierarchy
Time complexity
part
influence
depth
influence
Searching
Has parameter
procedure
Used for
In-order traversal
similarity
tree
Contrast to
Graph
Special operation
Has special type
Has special type
character
traversal
pre-order traversal
Has special type
evaluation
Binary Tree
functional
Has special type
Has special type
Has special type
post-order traversal
Almost completed binary tree
Strictly Binary tree
Has special type
Complete binary tree
28
Another example
Peirce, 1870
develop
General algebra Of relations
influence
John F. Sowa, 1984
DB design
Expert system
hierarchy
Quilan, 1969
NLP
invent
IR
part
invent
Applied in
Applied in
influence
Semantic network
Applied in
Applied in
procedure
A special type of
A system of logic
similarity
Is a
First Logic formulas
Conceptual Graph
support
knowledge representation
Used in
character
Is a
Conceptual schema language
Express
evaluation
Composed of
Semantic meaning
Composed of
function
Relation nodes
Has pros
Has parameter
Concept nodes
Has pros
type
Logically precise
Has parameter
Has pros
Has parameter
signature
valence
Humanly readable
Computational tractable
29
Link system tuning

• Purposes evaluate our link system and improve it
• T1
• 10-15 users
• Provide them with 50 pair of concepts identified
  as the import concepts in IT domain
• Ask them to name the link (either types or names)
  between concepts
• T2
• 10-15 users
• Ask them to create concept maps on the topics
  they are familiar with
• Evaluation
• T1
• link types or names are consistent between each
  pair of concepts among the users?
• compare with our link system
• T2
• whether there are new link types and link names
  used by at least 20 of the users
• Link system improvement

30
Agenda

• Introduction
• Research questions
• Literature review
• Proposed approaches
• Testbed and experiments design
• Timeline

31
Platform

• GetSmart Concept Map
• User functions
• Organize related urls, notes and images on the
  map
• Use semi-open link system to build concept map
• Link type users are required to select one from
  the list
• Link name users have 3 options
• Select a suggested link name under that link type
• Name it by themselves
• Leave it blank (unspecified)
• System functions
• Organize all the resources together
• Get statistic information of link type on maps

32
Hypotheses

• Research Question 1
• Rationale
• Our link system can help users to clarify the
  relationship
• Our link system can guide students view a topic
  from various aspects
• H1 Users who use our link system can build
  better concept maps than those who use open link
  systems
• H2 Users who use our link system can learn
  better than those who use open link systems

33
Hypotheses (cont)

• Research Question 2
• Rationale
• Each link type indicate understanding of one
  specific aspect of knowledge
• Instructors can provide instructions based on the
  statistical information of link types on the map
  using our link system
• H3. Users who get the instructions can build
  better concept maps those who use open link
  systems
• H4. Users who get the instructions can learn
  better than those who use open link systems and
  have no such instruction

34
Hypotheses (cont)

• Question 3
• Rationale
• Each link type(even link name) indicate one
  particular reasoning.
• We can match search tasks with link types and put
  these match into semi-structured queries, for
  example
• Give some examples of A Trace the link type
  hierarchy or even more specific link name is
  an example of and get results
• What causes A? Trace the link type sequence or
  even more specific link name causes and get
  results
• H5 From the knowledge repository which is
  integrated from all the concept maps using our
  link system, users who use semi-structured
  queries can retrieve information faster and more
  accurately than others who just use simple search

35
Metrics

• Evaluation of learning effect score of test
  designed to test the understanding of the covered
  topic
• Accuracy of searching score of test designed to
  searching tasks
• Evaluation of concept maps
• Compare to expert map
• Accuracy Are the concepts and relationships
  correct?
• Thoroughness Are important concepts missing? Are
  misconceptions apparent?
• Qualitative factor
• Organization Is it neat and orderly or is it
  chaotic and messy?
• Creativity Are there unusual elements that aid
  communication or stimulate interest without being
  distracting?

http//www.udel.edu/inst/jan2001/concept-mapping/
36
Experiment1

• Subjects
• 2 groups (G1 and G2, 10 in each group) of users
• Materials
• Require about 30 minutes to read
• About knowledge management
• Steps
• Pretest test on users on knowledge about A
• Users in G1 use an open link system, while users
  in G2 use our link system, creating concept maps
  independently
• All turn in their maps no longer than 1 and half
  hours
• Post test take exams about the materials they
  use after 1 week
• Evaluation
• Score the tests
• Evaluate concept maps

37
Experiment2

• Subjects
• 2 groups (G1 and G2, 10 in each group) of users
• Materials
• A topic covered by course
• Instructors map mainly include which kind of
  types user to address
• Tasks
• Every user create a concept map for a given topic
  in 10 days using our link system, submit the map
  on the fifth day and tenth day.
• On the fifth day, for users in G1, instructor get
  the statistical information about the link types
  every user in G1 uses, provide feedback to ask
  that user to explore the link types he lack
• Users take a test a week later on this topic
• Evaluation
• Evaluate the maps
• Score the tests

38
Experiment 3

• Preparation
• Select a group of concept maps related to a topic
  and integrate the resources related to maps and
  each nodes.
• Make these resource searchable
• Implement semi-structured queries
• Subjects and Tasks
• 2 groups (G1 and G2, 10 in each group) of users
• A test including 10 questions covered by the
  resources
• Users in G1 answer the questions by directly
  using maps , while users in G2 use map and the
  search function
• Evaluations
• Time
• Score the tests

39
Agenda

• Introduction
• Research questions
• Literature review
• Proposed approaches
• Testbed and experiments design
• Timeline

40
Timeline

• System design and proposal by 8/21
• Link system improvement 8/21- 8/28
• Tuning experiment and evaluation 8/28- 9/14
• Experiment1 and evaluation 9/14- 9/28
• Experiment2 and evaluation 9/28-10/11
• Experiment3 and evaluation 9/28-10/11
• Paper write-up 10/25