Mhairi McAlpine

1
Itembanking and Digital Repositories

• CETIS Metadata SIG
• HE Academy, York

2
Introduction and Definition

• This presentation looks at the requirements for a
  digital repository capable of supporting
  itembanking functionality.
• It suggests two ways that such a repository could
  be implimented taking into account the special
  requirements and data that items have.
• The itembank is considered to comprise of a store
  which holds the item together with its associated
  resources and a separate database which holds
  the metadata and related searchable information
  about the item. The itembaking system is
  considered to be the itembank plus the
  functionality required to enable it.

3
What does itembanking entail?

• Generation of items
• There needs to be some way of generating the
  items in an appropriate for entry into the bank
• Storage of items
• The items would have to be stored in a standard
  format.
• Item Description
• To facilitate re-use the accurate description of
  items is crucial
• Item Analysis
• Once you move away from a one use test,
  accurate analysis is crucial.
• Test Construction Algorithms
• These show how a test should be constructed from
  a bank,

4
What are the elements of an item?

• An item is considered to be the question and all
  of its associated information. Some of these
  will be metadata to be stored in a searchable
  database, but others may be resources to be
  stored with the items.
• Examples of Resources
• Base QTI file, which describes the item
• Called resource (such as an image, a video or
  animation) which appears with the question
• Stylesheet for embedded resources (such as
  MathML)
• (Response processing template?)
• Examples of Searchable information
• Metadata (UK Lom Core etc)
• QTI Metadata (additional metadata which describes
  items)
• Usage Data

5
Particular Features of Note about Items

• Item analysis
• statistical process(es) that items undergo to
  determine their quality and properties
• Item Usage Data
• output of analysis, based on the statistical
  processes undergone which have been defined in
  the glossary.
• Glossary (defined in QTI 2.0)
• A vdex file which describes item statistics

6
Overview of InfrastructureModel 1
7
Overview of InfrastructureModel 2
Test construction software
8
Advantages of piecemeal development through
unpacking functionality

• Not monolithic entity
• Thus easily adapted if model is wrong or if
  workflow processes are changed
• Can be built incrementally
• With a modular approach small chunks can be
  built and used immediately with existing
  processes used to fill in the gaps until the next
  pieces are built
• Can be upgraded incrementally
• As the specifications grow and develop, pieces
  can be upgraded in line. Furthermore it can be
  developed cross-institutionally ensuring
  community involvement.
• Easy to add new functionality
• As CAA beds in and people become more
  sophisticated in the use of itembanking
  additional demands will be placed on the system.
  A modular architecture allows
• for these to be slotted in at the appropriate
  points.

9
Consequences of unpacked functionality

• Need to ensure standards complience
• Not just to the strict specifications, but where
  the specifications are loose to ensure that the
  manner implemented is in line with existing or
  best practise.
• Need to live with a mixed model for some time
• As a workable system will not be developed in
  one go, manual processes and pre-existing
  software systems will have to be built into the
  workflow, requiring functionality that will not
  be required once the full system is in place
• Need for robust system testing
• Where a piece of the architecture is faulty, the
  possibilities for computer interaction with
  little user input may lead to difficulties with
  early detection of errors. System testing of
  each piece must be highly robust before it goes
  live.

10
Implications for Digital Repositories

• Security is a key concern in itembanking, direct
  human access to the bank is not considered
  desirable
• There are three lots of item information which
  people will want to select on metadata, QTI
  metadata and usage data
• Possible that usage data may not be held in the
  bank but generated at run time to get round
  issues of dynamic calibration.