What would be REALLY convenient is if the ISBN's of the related works could come with some bibliographic details. Title, author, maybe publisher and year. So I wouldn't have to look them up again myself. Not sure if that would be inconvenient to implement on your end though.

That MOSAIC thing, are you sharing your data with that project, then? Is there a way for a non-UK library to participate?

I don't think the project web site has gone live yet, but the JISC funded MOSAIC Project (an extension of the TILE Project) is working with several UK libraries to collate and mine circulation data.

The ISBN call is a little slow at the moment, because I'm using a text file containing ISBN to (bibliographic) ID mappings. I'll do some coding to turn the script into a mod_perl script that can load those mappings into memory.

For Umlaut, would it be useful to have an extra URL parameter that tells the script just to return the suggestions?

We don't currently make use of OCLC/LoC numbers, but I'll check to see if any of our MARC records contain them.

Do you use OCLC numbers or US Library of Congress numbers in your catalog? If so, it would be awesome if those could be used as access points too.

You realize the next step, right? Figure out a way for someone else who can figure out how to track usage in their catalog to send you data in a standard format, so you can include data from other libraries as input into your grouping algorithms too.