Metadata Management and Learning Object Composition in a Self eLearning Network

1
Metadata Management and Learning Object
Compositionin a Self eLearning Network

• Nicolas Spyratos
• spyratos_at_lri.fr
• Joint work with Philippe Rigaux
• Laboratoire de Recherche en Informatique
• Universite Paris-Sud
• Centre dOrsay
• France

2
overview of my talk

• motivation, context, objectives
• key issues
• learning object creation and metadata
  management
• searching for objects of interest
• work in progress / perspectives

3
motivation

• the facts
• traditional learning is premised on a paper
  medium that is difficult to
• produce/distribute/archive/duplicate
• e-learning is premised on the electronic medium
  which shares none of these
• features and thus naturally facilitates the
  learning process
• the (inevitable) conclusion
• use the Web for
• collaborative creation, distribution,
  archiving, sharing
• of learning resources

4
context

• EU IST Project SeLeNe Self eLearning Networks
• we view a SeLeNe as a network of
  authors/learners, collaboratively creating
• and using learning objects (LOs)
• one definition of learning object
• electronic, sharable chunk of reusable learning
  content, available on the Web
• (Simon and Quemada, 2002)
• The Esp_at_don project course-sharing
• (University of French Polynesia)

5
Self eLearning Network
module
INTERNET
PEER
PEER
ppt component
PEER
lesson
CSCourses
PEER
AICourses
6
the driving ideas

• performance
• response times should not exceed a certain limit
• simplicity
• creation/learning processes should be automated
  as much as possible
• LOs reside at the repositories of their creators
  (no moving around)
• flexibility
• communities of various types/structure should be
  easy to form
• joining/leaving a community should be easy
• uniformity
• authors and users should be treated as peers

7
Key Issues / System Requirements

• creation/management of learning object metadata
• ? automated as much as possible
• searching for LOs of interest based on metadata
• ? automated as much as possible

8
learning object metadata
LO id metadata content
? descriptions are
given using terms from a commonly accepted
terminology ? the challenging part is description
of content and its automatic creation
not considered in this work
LO content
description of
other (language, author, ..)
an address where the LO can be accessed
several metadata standards already exist today
(ex IEEE LOM) their role is to enable
metadata-based exchange, reuse and search of LOs
9
description of content(informally)
to describe the content of a LO collect
together the descriptions of its parts in a
manner that reflects its structure (cf table
of contents in a traditional textbook) LOs are
built up from atomic LOs in various
ways atomic LO any identifiable piece of
learning material its granularity is up to its
author composite LO consists of a set of LOs
that can be atomic or composite this set can be
unstructured, or structured in one of several
ways the issue generate content description
automatically, based on the descriptions of atoms
10
description of content (formally)
a LO is seen as an identifier o associated with a
set of LOs, called its parts parts(o) o1, ,
on if parts(o) ? then o is called atomic else
o is called composite components of a
LO if o is atomic then comp(o) ? else comp(o)
parts(o)?comp(o1)? ?comp(on)
O
O1
O2
O21
O22
O11
O12
11
description of content (LO structure)
? if we assume the following no-redundancy
constraints - no node can be a component of
itself (i.e., all nodes are different) - every
node (except the root) is part of one and only
one node then the LO structure is a tree
with the atomic LOs at the leaves
12

• description of content (taxonomy of terms)
• the description of LO content is a conjunction of
  terms
• from a commonly accepted taxonomy
• a taxonomy is a pair (T, ) where
• T is a set of keywords or terms, called the
  terminology
• is a reflexive and transitive relation over T,
  called a subsumption
• terms s and t are synonyms, denoted s?t, iff st
  and ts
• several standard taxonomies of topics already
  exist today
• (ACM, IEEE, most of them being
  tree-taxonomies)

13

• Example of taxonomy
• (fragment of the ACM Computing Classification
  Scheme)

Programming
Algorithms
Theory
Languages
Sorting
OOL
Merge
Quick
Bubble
C
Java
JavaBean
JSP
14

• description of content (labelling the
  composition tree)
• atomic LO
• the description is a conjunction of terms chosen
  by the author
• providing a description is mandatory (when
  registering the object)
• composite LO
• form the disjunction of the descriptions of the
  parts
• transform to CNF
• replace each disjunct by its lub
• minimize, i.e., eliminate implied lubs
• ? the result is the most precise description that
  we can make of the LO content

15
O sorting
O Java
O1 Java ? Quick
O2 Bubble
O1 JavaBean
O2 JSP
disj
disj
(Java ? Quick) ? Bubble
JavaBean ? JSP
CNF
lub
(Java ? Bubble) ? (Quick ? Bubble)
Java
lub
?
programming
sorting
min
sorting
16

• Example of taxonomy
• (fragment of the ACM Computing Classification
  Scheme)

Programming
Algorithms
Theory
Languages
Sorting
OOL
Merge
Quick
Bubble
C
Java
JavaBean
JSP
17

• description of content (composite LO)
• the author of a composite LO may provide own
  description, to indicate
• possible uses of the LO, other than those
  inferred from its parts
• composite LO
• infer content description from descriptions of
  the parts (as before)
• form the conjunction with authors description
  (if any)
• ? intuitively, in giving a description, the
  author treats the composite LO as an atomic one,
  adding on to the knowledge already built about
  the content

18
GeomShape ? JapFlag
author input
JapFlag
GeomShape
description derivation
Rectangle ? Circle
O
Circle
Rectangle
19

• description of content (other derived metadata)
• several useful metadata can be derived
  automatically
• from content description
• table of contents
• index
• catalogue

20
description of content (table of contents)
the table of contents being a linear
presentation of the components, every possible
traversal of the composition tree provides a
table of contents
O
O d O1 d1 O11 d11 O12 d12
O2 d2 O21 d21 O22 d22
LO
O1
O2
O11
O12
O21
O22
its inorder table of contents
21

• description of content (index)
• LO index
• the set of all pairs (t o1, o2, , on) such
  that
• o1, o2, , on are the LO components in whose
  description t appears
• (its the index as we know it from traditional
  textbooks)
• the index can be computed from the LO tree

22

• description of content (catalogue)
• LO catalogue
• the set of all pairs (t, o) such that t appears
  in the description of o
• (its a sort of shopping list)
• the catalogue can be computed from the LO tree
• the concept of LO catalogue is the basic concept
  in implementing
• searching for LOs of interest in the network
• the union of catalogues of all registered LOs is
  the network catalogue

23
Searching for learning objects (the network
catalogue)

• - the network catalogue C(term, id) relates terms
  to registered LOs
• the terms of the catalogue are structured by a
  subsumption relation

a
b
terms
c
d
f
g
h
e
LOs
1 2 3 4 5 6 7 8
extension of a term t Ext(t) o (t,
o)?C ? the catalogue taxonomy allows for
browsing and querying
24

• searching for learning objects
• (browsing the network catalogue)
• show top terms then continue with the successors
  of those selected by the user

a b c d e f g h
ex if c and d are selected then we continue
with f, g and h
25

• Searching for learning objects
• (querying the network catalogue)
• searching for LOs indexed by a term or a
  combination of terms
• a query is a boolean combination of terms
• q t q1?q2 q1?q2 q1 ? ?q2 ?
• its answer is defined recursively as follows
• ans(t) Ext(t) ? Ext(t1) ? ? Ext(tn)
• where t1, , tn are the immediate
  successors of t in the subsumption
• ans(q) if q t then ans(t)
• else begin if q q1?q2 then
  ans(q) ans(q1) ? ans(q2) if q q1?q2
  then ans(q) ans(q1) ? ans(q2)
• if q q1? ?q2 then ans(q) ans(q1) \
  ans(q2)
• end
• ans(?) ?

26
Searching for learning objects (maintenance of
the network catalogue)
to register a LO add its catalogue to the
network catalogue NetCat NetCat ?
LOCat to unregister substract its catalogue
from the network catalogue NetCat NetCat \
LOCat to modify a registered LO add the new
pairs and remove the old NetCat (NetCat
\ Old-LOCat) ? New-LOCat ? conceptually,
maintenance is straightforward
27
other network services
registration making the catalogue of a LO
available to the network catalogue (or removing
it) change propagation notification of change
to users of LOs user profiles in the form of
queries (for notification of new LO creation)
28
to summarize

• each user (author, learner) creates LOs and
  stores them in a local repository
• to make LOs sharable in the network their owners
  must register them with the network catalogue
  and inform of any changes later on
• content description and related metadata can be
  generated automatically from the network
  catalogue
• searching for LOs of interest is done by
  browsing/querying the network catalogue
• a number of other services can be provided based
  on the network catalogue
• conceptually, catalogue maintenance is
  straightforward
• designing a generic module for the management of
  network catalogue provides flexibility in forming
  user communities

29
Ongoing work
case study metadata management in XML to
experiment with what we have so
far personalization views of the network
catalogue
30
perspectives

• relations other than parts-of for creating
  composite LOs
• ex prerequite-for relation over LOs (cf
  dependency graph in a textbook)
• learning trails
• sequence of LOs in a prerequite-for relation
  forming a learning course
• any subsequence of a topological sort in a LO
  is a possible learning trail
• the LOs of a learning trail may not belong to
  the same composite LO
• i.e., the prerequite-for relation doesnt
  depend on the parts-of relation
• learning scenarios passive, active, nomadic,
  peer-to-peer
• a learning trail may be considered as a
  component of a composite LO
• ? use of RDF to express such relations (or
  some other formalism)
• catalogues with different taxonomies
• ? need for semantic mappings (e.g., between ACM
  and IEEE)

31
example of dependency graph

• a useful information on how to use a composite LO
  during learning is knowledge of the dependencies
  among its components
• PLs Algos
• Pascal Java Lisp
  Quick Bubble
• QuickJava BubbleLisp
• this prerequisite-for relation over the
  components of a composite LO is what we call a
  dependency graph
• may be given by the author
• several dependency graphs may exist
• can serve as a basis for defining learning
  trails