Real-World Implementation Report

[Eliot Kimber <ekimber@reallysi.com>]

Thought I would report on my experience so far using the 8 March 
Learning specialization as a base for my test-prep publisher's question 
content.

So far it's worked remarkably well. Here is a typical instance of my 
specialized markup:

<question>
   <p>This is a paragraph before the question proper.</p>
   <multiple-choice>
     <prompt>What is the volume of this figure?<art><image
           href="art/93OH_PT1_A01.eps">
           <alt>This is the alternative text for the figure</alt>
         </image></art>
     </prompt>
     <mc_answer>
       <answer_option>
         <answer_text>18 cubic centimeters</answer_text>
       </answer_option>
       <answer_option>
         <answer_text>36 cubic centimeters</answer_text>
       </answer_option>
       <answer_option>
         <answer_text>48 cubic centimeters</answer_text>
       </answer_option>
       <answer_option><answer_text>54 cubic centimeters</answer_text>
         <correct_response/>
         <feedback>Width times height times length: 3 x 3 x 
6</feedback></answer_option>
     </mc_answer>
   </multiple-choice>
</question>

Here <question> is a specialization of "container" (new in DITA 1.2) 
that serves to bind the interaction to any lead-in blocks and to 
classifying metadata.

<multiple-choice> is a specialization of lcSingleSelect and the mapping 
of the subelements should be clear enough. I specialized primarily to 
provide more familiar/friendlier element type names, but I would expect 
most uses of the learning specialization to do the same.

The <art> element is just a ph specialization that binds one or more 
<image> elements to classifying metadata (for this client art instances 
are "managable objects" in the CMS that would be used by reference).

One aspect of this markup approach is that I largely avoided the need 
for multi-paragraph questions, although I am running into some bad 
examples in tests that emulate state-specific tests, such as this question:

<question
   outputclass="boxed">
   <p>No one lives his or her life alone. We all have families. We all 
have friends, or at
     least, people who come and go in our lives. Some people have 
influence over us, some
     people do not.</p>
   <p>Write a composition about someone (a family member, a friend, even 
a stranger) who has
     had some influence, positive or negative, on your life.</p>
   <p>Before you begin to write, think about what force of this 
individual’s personality
     might have had an influence on the way you act or see the world.</p>
   <open-ended>
     <prompt>
       <b>Now write a multiparagraph composition about a specific moment 
or moments where
         things might have changed for you, where you had a realization, 
or came to a
         decision thanks to that person. Write your answer on a separate 
sheet of paper.</b>
     </prompt>
     <open_answer/>
   </open-ended>
</question>

If you read the question you'll see that it's really two questions 
("write a composition" and "write a mutli-paragraph composition") but in 
the test it's presented as a single question. The markup here is not an 
exactly correct semantic representation of the original question.

It might be better, for example, to have my <question> element allow 
multiple "open-ended" elements, which would allow me to more accurately 
reflect the intent of the test.

Unfortunately for this client they have to emulate what a given state 
does--they can't appeal to good practice or necessarily apply editorial 
authority.

I defined a few generic topic types to organize the publication content: 
test, question_set, etc. and have a specialization of section called 
"question_group" that allows me to have groups of related questions that 
might have a title. These all allow me to model the organizational 
structure and content of the print publication. My "question_set" topic 
type can be used within the context of a publication to reflect 
subdivisions or as a standalone topic that serves as a pool of one or 
more re-usable questions.

As part of the project for this client we are managing questions as 
individual objects in the CMS with the goal of building a simple 
question management system that will focus on being able to search for 
questions based on their classification within both subject taxonomies 
as well as with respect to state-specific learning standards.

Our general business goal is to generalize this functionality within our 
RSuite product so we can provide some general out-of-the-box learning 
content support features. These will in turn leverage our general DITA 
support features, so having the Learning specialization is of tremendous 
value to us as service and software providers as well as of obvious 
value for our current and potential clients.

For a level of effort benchmark, I started on this last Friday morning 
and spent about 8 hours over the weekend coming up to speed on the 
Learning specialization, including taking the time to provide some 
feedback, and so on, and then created sample docs and refined my 
specializations iteratively using OxygenXML as my development 
environment. By the end of the day yesterday I had a pretty solid markup 
design and representative sample docs reflecting real published test and 
prep lessons. Most of that time was spent working out the details of my 
higher-level organizing structures. The actual question markup was very 
easy to design and implement from the Learning specialization once I 
understood what the general pattern was.

I spent a couple of hours today building an HTML preview transform for 
use from our CMS product but I'd need to spend about the same amount of 
work extending the Toolkit's HTML process to support specific 
specializations. (and could largely reuse the code I wrote for my 
preview script.)

So so far so good. This client doesn't really do anything with 
objectives or other aspects of learning content--the focus is almost 
entirely on questions. Right now their business focus is on improving 
their print workflows but they do have a medium-term interest in more 
electronic deliverables so I'm hopeful that the Learning base will pay 
dividends there.

Cheers,

Eliot

-- 
Eliot Kimber
Senior Solutions Architect
"Bringing Strategy, Content, and Technology Together"
Main: 610.631.6770
www.reallysi.com
www.rsuitecms.com