[QTI] Re: Hi and some questions

[Michael Piotrowski]

On 2007-09-12, Niall Barr <niall at nbsoftware.com> wrote:

>> this is because QTI 2.x is extremely complex and in many cases not
>> sufficiently specified.  So you have to put in a lot of work to get no
>> or only very limited interoperability.
>> 
> QTI 2.x is much more clearly specified than 1.x was, but it is complex
> because it aims to cover a great variety of requirements.

Yes, I agree.  However, it is (IMHO) debatable, whether this is the
right approach.  QTI 2.0 certainly can be used to describe a wide
variety of assessment items; this flexibility comes with a price,
though: Since there are so many possibilities for both encoding and
interpreting items, interoperability is hard to achieve.

> There is a subset of 2.0 (using the predefined template processing)
> which is fairly simple to implement, and also it is usually fairly
> easy to get answers from the authors where the specification lacks
> clarity :-) .

While I appreciate this, it is, unfortunately, not the same as a formal
specification.

> There probably is a need for a much larger set of example items for
> all the specifications to help developers with interoperability, but
> generating and documenting good test items takes both a lot of time
> and a detailed knowledge of the intricacies of the specification. (I'd
> be happy to do it though, if someone will pay for my time.)

Examples are helpful, but they are no replacement for a formal
specification.  The QTI 2.0 specification describes too many
requirements, assumptions, etc. only in informal language, and does not
formalize them in the XML schema.  This makes validation very hard.

Greetings

-- 
Michael Piotrowski, M.A. <mxp at ifi.uzh.ch>
Institute of Computational Linguistics, University of Zurich
OpenPGP public key ID 0x1614A044