"The Sponsor" has a Bright Idea — Exploratory Concept Phase

During this phase, the primary requirement is simply to explore the idea. In this case the idea was to reveal the breadth of interpretation of commonly-used project management terms, and the importance of addressing this communications obstacle.  The data gathering started as simply the path of least resistance:  an MS Word document.

The Idea Gathers Steam — and Friction

The idea starts to show some value — "Yes, indeed there is a breadth of interpretation and it is worth discussing".  But two things usually happen at this point.

• As the amount of data increases, the original low-structure repository becomes unwieldy to maintain and grow.  (You don't want to edit long tables in MS Word).
• New uses for the data are imagined, which the original repository is unsuited to feed. (Maybe web-pages would serve users better than a PDF?  And what about hyperlinks?)

Infuse Structure -- Development Phase

To leverage it better, the heretofore unstructured data must be corralled into a more structured form.  MS Access was selected as a convenient database, and we created some software in Delphi to parse the Word document (by that time 2000 definitions) into Access.

As usually happens when infusing structure that had not previously been enforced, this required a few iterations of the software to recognize all the nuances in the document, and finally a few revisions of the source document to fix up inconsistencies in the manually entered data.   One guiding idea at this stage was to avoid hand-editing the data in the database (until we were absolutely sure we were ready to abandon the source data) as any database changes would get overwritten if another parsing pass was needed.

Although we didn't need to preserve most of the existing hand-applied font and character-style attributes, in fact a small proportion of entries had useful formatting, such as list-structure.  To accommodate these, we decided to export from Word in HTML format, eliminate unwanted formatting during parsing, and store all entries in the database in HTML to capture the remaining desirable formatting..

Implementation and Production/Maintenance Phases

Clearly this is not a very large IT project, barely worthy of the terms like "Production".   Yet even this size of project benefits greatly from using a database as the information repository and structuring vehicle ...

With the data safely stored in Access, we can perform a variety of convenient manual operations, and we have knocked together a handy definitions editor which allows us to view the HTML results as we enter or edit an entry.  Pages are automatically generated from the database in a fairly obvious fashion, with some attention paid to breaking the pages at reasonable boundaries. 

The trickiest (and most speculative) part is the hyperlink find-and-insert mechanism.  In essence it hunts through each definition looking for occurrences of the terms elsewhere listed in the database ... and inserts a hyperlink.  However, a few extra heuristics (not at all resembling artificial intelligence) are thrown in to try to avoid unreasonable links.   A few silly links still manage to creep in, but the result is surprisingly good for a completely hands-off operation that eliminates an otherwise prohibitively laborious task.

Graham Wideman
Home Page: Wideman-One