Kaboodle Bookmarklet - Method?

Hey,

I was just wondering if anyone had a more in depth concept on how the Kaboodle Bookmarklet works or how to mimic the concept.

I know there would be some sore of page scraping. I've tried the Kaboodle bookmarklet on many different pages and not all pages have descriptive identifiers in the content or the class's/id's the descriptions sit in. Yet in can still manage to find very relevant content.

Suggestions or ideas ?