nodalpoint.org

I've never found anything that does what I want (i.e. make sense of several hundred PDFs in a directory with arbitrary names ;-) ). Problem is, high-throughput data people like us tend to save files first, assume there's a solution later. There are 2 basic, obvious problems: (1) PDF format, which is neither marked up nor parseable, (2) no direct link between a PubMed ref and its corresponding PDF. I think the only solution is to name those PDFs as others have suggested.
There are a couple of tools I like. One is RefDB. It comes with scripts that convert Medline to RIS, then you can import RIS to MySQL. There is also Librarian, though this requires a lot of manual operation, reference by reference. But in both cases, I like the idea of a shared repository. Also there is PDFSearch, which essentially converts PDF to plain text and allows searching. I keep thinking I should be able to pattern match fields from my RefDB database to text from the PDFs and add an AV: field to the RIS record if it all matches, but haven't got around to it.
Literature management is an area I wish I had more organised, but at the end of the day I spend more time doing data analysis than reading/writing papers. It's an area ripe for faffing about, procrastination and displacement activity.