Since we saw Google Suggest, Google Maps and GMail we've done some experiments too. One of the things was to take all the controlled headings from WorldCat and do a suggest-like service. Works pretty well, and it might be something we'll pursue later.
Lately, though, I've been applying AJAX to a Dewey browser. This was inspired by the OCLC Research Software Contest. Two of the things we have available are all the records with 'Ohio' in them from WorldCat and the DDC Summaries. So I tried my hand at YADB (yet-another-DDC-browser). There are a lot of them out on the web, but I had the idea of using the Rocks/Ganglia visualization approach to help guide browsing. For various reasons we ended up moving away from the 10x10 display to successive rows of 10, and in the process have gone through a half-dozen different approaches. We think we've got one now that's going to work, though.
Another technique you'll see used is to load data into HTML iframes. The nice thing about doing this is that the browser's history gets updated (so the back button works), and if you're loading XML with an associated XSLT file, that XSLT transformation happens automatically.
The real trick to the system, though, is what happens in the server. We've gone through several iterations there before we found a simple scalable solution. Overall, the lines of code involved in the system is probably a third of where we started. More on what I think of as our 3-level server side architecture in another post.
Thanks to Diane Vizine-Goetz (our DDC expert here in OR) for all her help with the browser. In addition to browsing a collection of records, Diane thought of loading the DDC schedules themselves and using it for classification. A sample of the Abridged DDC is show in the screen shot above. Another aspect of interface design is making the whole thing pretty, simple, and functional, all at the same time. Lance Osborne did most of the design work.