Transcript Learning Object Metadata

Learning Object Metadata
From the locally prescribed to the socially derived
(or, a look back at 4 years of LORNET at the University of Saskatchewan
Scott Bateman, Christopher Brooks, Gordon McCalla, Jim Greer
Advanced Research in Intelligent Educational Systems
Department of Computer Science
University of Saskatchewan, Canada
In the beginning...
ARIES Laboratory
University of Saskatchewan
 The world of Learning Objects was still in
early definitional stages
 The focus was on metaphors of LEGO, atoms,
organic structures...
 But metaphors can be quite limiting
 They tend to limit your vision
 Are learning objects just like physical
structures? Not really...
 Metaphors ignored the issue of metadata,
and focused on the “physical” connections
between learning objects
In the beginning..
 Metadata was simple
ARIES Laboratory
University of Saskatchewan
 Text only, mostly following the Dublin Core
specification of authorship, subject, and
keywords
 No ability to describe intended usage
 Really, there are two kinds of intended usage
 how the system should use the learning objects
 and how the users did use the learning objects
 Metadata was generated by the authors of the
content directly, and was laboursome to create
due to lack of tools
 Instructors still needed to manually find and
aggregate resources into courses...
Contributions to LO Metadata
ARIES Laboratory
University of Saskatchewan
 Models for learning object metadata that
make use of AI techniques
 Ontologies for describing what students have
done (cf. LOCO Analyst by Jovanovic et al.)
 Ontologies for describing how the domain
connects to learning
 High level approaches for metadata
creation
 Agile User Modelling (Vassileva et al)
 Ecological Approach (McCalla)
 Flexible Learning Object Metadata (Brooks &
McCalla)
Specific Systems
 iHelp Courses
ARIES Laboratory
University of Saskatchewan
 Is able to provide simple sequencing adaptivity
based on end-user data
 Provides a methods for determining what
students have done
 Provides privacy for the protection of users
when data mining and visualization activities
occur
 MUMS
 Architecture of a distributed user modelling
system, providing SOAP-based subscription
access to RDF knowledge
 Knowledge can be traditional metadata, or
metadata fitting new schemas
Specific Systems
ARIES Laboratory
University of Saskatchewan
 Open Annotation and Tagging System
(OATS)
 Social construction of learning object metadata
through collaborative tagging
 Leverage the “note taking” approach of
students
 Recollect
 Video course casting system
 Metadata created through data mining slide
images
 Variety of techniques used, including conversion of
images to textual data, pixel and block characteristics,
etc.
Questions
ARIES Laboratory
University of Saskatchewan
 As usual, we are left with more questions
than answers
 Is metadata a collection of information, or a
process of deciding over the usage knowledge
that has been collected?
 Is automated metadata generation from
content artifacts a reasonable approach?
 Experiments suggest that students, instructors, and
automatic metadata generation tools all classify
content differently (Bateman et al., WWW 07)
 Where did all the agents go?
 Metadata is still just human text, or very limited
vocabularies...
Important Directions
ARIES Laboratory
University of Saskatchewan
 Humans give us a very important kind of
metadata just by using our systems
 Content metadata, such as in collaborative
tagging systems, is great for exposing to end
users who are looking for informationAttention
Metadata
 Pivot browsing, association creation, and categorization
using collective intelligence
 Usage metadata (Attention Metadata) helps
decide how to change things automatically
 Automatic capture of learner interaction data, possibly
mined and interpreted and attached to learning objects
 Content adaptation, recommendation, etc.
 Learning environments need to be smart, and share
how people interact
Thank You
ARIES Laboratory
University of Saskatchewan
• Thank to our many collaborators on this
work
• This work was funded principally by
NSERC through the LORNET Network
• ai.usask.ca