Wednesday, March 30, 2005

Yes! I am the "worst"!'s Best of the Web Directory picked as a Forbes Favorite in their Special Needs Resources section. But this part obviously caught my eye:
WORST: The unwieldy general message board would benefit from splitting into subsections devoted to specific issues.
I can't deny it. We don't have any subsections on our message board. Right now, it feels that any attempt to explain how this came about will sound like pure defensiveness. Instead, let me simply say that it was a conscious attempt to not succumb to a major pet peeve I have with message boards (as they currently stand). As Lee LeFever has pointed out, message boards are good Flows, but poor Stocks:

The question is: How can we ensure that flows are being stocked effectively? What makes a good stock?

Websites that flow will find that the management of stocks becomes very important over time. As the flows create new information, the stocks grow to become a rich, but often unwieldy resource. For this reason, website owners should pay particular attention to how stocks are being archived for reference. Easily accessible stocks compliment engaging flows.

During the process of parents connecting to support each other, there is a flow of valuable support in both emotional support and informational support. However, getting that valuable flow that back out is the key problem. It gets complicated by the fact that (for me) the information isn't neatly categorizable. School issues blend into home issues. Attention issues blend into reading issues. Even distinct Federal Education laws (Section 504 and IDEA) touch on similar issues (specifically classroom accommodations). Right now, there is a lot of benefit from parents interacting with a diverse range of parents. So how do I keep the discussions flowing, but derive some kind of sense to the stock that is four years worth of parents helping each other?

At the beginning of March, I knew that some kind of bottom-up categorization was what I needed. "I have all this data, there must be some way to highlight specific topics so people can dive in an swim around in the areas they want while having seamless access to the other parts." It will look like subcategories, but it just be one facet of the large complicated jewel that is the world of learning disabilities. Once I got into the idea of thinking in terms of categories as a property of messages instead of categories as buckets for messages, I thought my solution was meta-tagging or some kind of folksonomy. I did my best to keep up with the debate around folksonomies and I was thinking up complicated ways to address the "people are bad categorizers" and "people are lazy" problems.

At this time, I attended the NHVA Innovations Conference. Actually I spoke and, good gravy, they have an mp3 of it (I'm not linking that on purpose). There I met Peter Kaminski and he gave me a new perspective. He pointed out that our messages have all the contextual data I needed. What seemed more appropriate was a search tool that would group the data together by similarities much the way Google News is able to group similar news stories based on the text in the story itself. Man, talk about a *phoom* moment. Thanks, Peter!

What tool would that be? Vector space searching. I'm learning about it and here's where I'm starting:

Building a keyword vector space engine in ColdFusion: adding an extra dimension to your keyword searches

Building a Vector Space Search Engine in Perl

It's not that I think metatagging is right out. I just realize now that with large amounts of text data, more hand-entered text data probably isn't the right direction.