I'm glad to see a discussion of typed citations happening here. I find that it's one of the first questions I get lots of times when I discuss altmetrics.
Citation typing or semantic citation is a bit different from a lot of the other work being done under the "altmetrics" banner in that there is a very long history of work in this area (unlike, say, mining Twitter citation), from a number of different directions.
Toulmin's model of argument, which establishes a taxonomy of link types (refutes, agrees, etc) dates back to the late 50s, and has I think been used a lot in legal education, for analyzing decisions. Ted Nelson was working on typed hyperlinks in the 60's with Xanadu, and more recently the "argument interchange format" has been proposed to link networks of claims and counterclaims across the Web. The Hypothes.is project is working to do the kind of crowdsourced annotation of arguments that Fred mentions, and while it's IMHO kind of a long shot, it has plenty of potential and awesome implications. Wikimedia's Wikicite project, another crowdsourced citation database, may or may not support typed citations, but it certainly could.
In the academic sphere, Simon Buckingham-Shum has been working on ontologies of scholarly argument for decades, and has created the ScholOnto ontology as well as many annotation tools. Lighter-weight attempts have included things like Tree Trellis and Project Archelogos, which used simpler ontologies, sacrificing descriptive power for simplicity and ease of use.
This latter feature is a key concern in these typing systems, because one of the biggest impediment to their use is what Buckingham-Shum calls the "capture bottleneck:" the difficulty of converting (or converting authors to convert) unstructured text into semantic markup. Indeed, this is a problem for the whole semantic-web enterprise (as Cory Doctorow notes in his wonderfully prescient 2001 essay "metacrap" http://www.well.com/~doctorow/metacrap.htm#2.2
). Crowdsourcing may be one answer, but it's far from a magic bullet (it's a naturalistic bullet?)...projects for crowdsourced annotation have certainly failed more often than they've succeeded.
Sorry if this rambles a bit, or if you're familiar with all these already. I think it's worth getting out there, because discussions of typed citations often seem to lack some of the historical and cross-domain contexts that they'd really benefit from; I don't think it's always appreciated that this is a Hard Problem.
That said, it's also clearly one that's worth working on, given the potential payoffs. Vanever Bush imagined the Memex as being able to traverse and navigate webs of ideas in a much more nuanced way than we can do with the current Web...it would be awesome if we could finally make it happen.
UNC Royster Scholar
School of Information and Library Science
University of North Carolina at Chapel Hill