Reflections so Far

1.R2Q2 was tricky to set up, however they've updated their Installation Tool (UPDATED 13/3/07) so this may be simpler now. But once it was working and I figured out how to use it (they also seem to have added help since), it was fairly simple. However it is only possible to load one question at a time and questions need to be in a zip file, thus making its use limited and adding questions one at a time a bit cumbursome, this would also have implications later.

2.In QM perception 4 it is possible to load a number of questions together as a quiz. However QM Perception is more complex than R2Q2 which allows it to have more functionality but also makes it harder to figure out (not being a great one for reading unstructions).

3.The knock on effect to development of our mcqfm project was a requirement for single questions in QTI for R2Q2. We considered outputting QTI as single questions but this would mean restricting the user at the interface to one question at a time and making the development of QTI questions one at a time rather more tiresome. In the end we have opted to add a surplusWrapperNode as a root Node to encapsulate the QTI questions as a quiz. The surplusWrapperNode being a temporary expedient at the proof of concept stage, with the idea that in the future a further operation of removing this surplusWrapperNode and zipping up will be required before the questions can be usefully used by r2q2.
Probably what we might do by the end of the project is output the lot as an in qti 2.1 terms, then the new tools which are being developed could display that.

4.Development of the QML for QM Perception was relatively straightforward, as I've said already they allow for a quiz thus making it possible to generate a text stream which is immediately importable into perception.

5.In the BQL (the Basic Question Language - we use as an intermediate notation between the QTEXT humanly readable notation and the full XML -QML/QTI- representation) we wanted to generate XML that was expressive but simple, an extension of the humanly readable QText. It could be expressive in its use of elements, e.g , and attributes e.g scorevalue, correctorder etc. while still retaining readability.An example question for the order model is as follows:

<ns:item model="order" id="2">

<ns:orderquestion totalscorevalue="1" minvalue="1">

<ns:content>What is the correct order of the following notes:</ns:content>

<ns:orderanswer correctorder="3" id="1" correct="true" scorevalue="1">me</ns:orderanswer>

<ns:orderanswer correctorder="1" id="2" correct="true" scorevalue="1">doh</ns:orderanswer>

<ns:orderanswer correctorder="2" id="3" correct="true" scorevalue="1">

ray</ns:orderanswer>

<ns:feedback>The correct order is: doh, ray me</ns:feedback>

</ns:orderquestion>

</ns:item>

It might be nice to add scoring for questions, adding differential scoring for different responses in mcq or cloze questions is certainly a possibility - but would that complicate the QText to such an extent that its fundamental goal - to allow *anyone* to produce questions in QTI would become compromised (since it would require more recondite knowledge and therefore become only usable by people who know the spec).

6.Further questions on R2Q2 that might be represented by QText are inline choice, associate, match, gapmatch and hottext and mathematical template questions.

by Caroline