use of contextualized attention metadata for ranking and recommending learning objects
DESCRIPTION
Presentation at CIKM 2006 about how to calculate the relevance of a learning object based on the attention that the object have receivedTRANSCRIPT
Use of Contextualized Attention Metadata for Ranking and
Recommending Learning Objects
Xavier Ochoa, ESPOL, Ecuador
Erik Duval, KULeuven, Belgium
Agenda
• What is the problem?
• Use of CAM for Ranking and Recommending
• Demo: ShareLOk
• Conclusions
Human Review
Human Review
TFIDF
What is the problem?
• A Ranking mechanism must be:
MEANINGFUL and SCALABLE
Ranking Strategies for LO
• There is a lot of ideas that could help to implement a Ranking Mechanism for Learning Objects
• BUT THE SEMANTICS ARE DIFERENT
Web pages Papers LO
• Eg. A link from one page to other means support. A link from one LO to other does not carry the same meaning.
Network Analysis
• There is not a linking structure between LO like in Web pages or Scientific papers
• But we have plenty of relations between different entities and LO– Authors, Instructors, Readers, Courses, LMS
• The main source of this information can be extracted from Contextualized Attention Metadata (CAM)
Lifecycle Actions Main Information Source
Creation Creatingauthor, components Authoring tool,
Components
Labeling Labelingmetadata format, origin, confidence
Authoring tool or Metadata generator
Offering Inserting inserter LOR or Sharing app.
Selecting
Searching query, results LOR’s search tool
Recommending objects recommended Recommender
Browsing Time LOR or Recommender
Selecting object identifier LOR or Recommender
Using
Publicating LMS context LMS
Sequencing list of sequenced objects ID tool or Packager
Viewing Time, tool used Browser or app.
Annotating rate or review LMS
Retaining Retaining decision to keep or delete LMS
Network Analysis
• CAM as K-Partite Graph
O 1
O 2
O 3
C 1
C 2
U 1 U 2
A 1
A 2
User Partition
Course Partition Author Partition
Object Partition
Network Analysis
• CAM as 2-Partite Graph (User-Object)
Network Analysis Metrics
• From this Graph we can create serveral metrics. For example:– Popularity Rank (PR)
– Author-Corrected Popularity Rank (ACPR)– Weighted Popularity Rank (WPR)– “Best-Sellers” Rank (BSR)
Similarity Metric
• The K-Partites Graphs generated from CAM can be folded into Normal Graphs
• These graphs have a link between two nodes in the same partition, if they have been linked to the same object in the folded partition.
Similarity Metric
U1
U2
U3
O1
O2
O3
U4
U5
U6
U1
U2
U3
U4
U6
U5
2-Partite Graph (User and Objects) Folded Normal Graph (Users)
Similarity Metric
• Object Similarity based on Downloads
• Object Similarity based on Re-Use
• Users similarity based on Downloads• Author similarity based on Re-Use of
Components
Object Similarity
Communities ARIADNE
Personalized Rank
• We can create a profile of the user based on its CAM
• We will use the same LOM record to store this profile
• Instead of having a crisp preference for a value, the user will have a fuzzy set with different degrees of “preference” for all the possible values.
Contextual Recommending
• If the CAM is considered not only as a source for historic data, but also as a continuous stream of contextualized attention information, we can use very recent CAM to generate recommendations based on what the user is focusing his/her attention at the moment.
Demo
Conclusions
• Meaningful and Scalable Ranking Metrics can be created for Learning Objects as has been done for other fields
• But first we need to measure, analyze and understand the data about Learning Objects… We need Learnometrics
• Development in Learnometrics could be easily translated in better tools
Thank You!
• Questions, Comments, Critics….
Are all welcome!!