I'm not following links at the moment, just pulling data out of the tweets themselves. I've had some success filterring spam out but intend to imropve that further as soon as I get a chance. There are some more fairly simple things I can do, but there's probably a limit to how good I can make that. As for the vagueness, I think that's in part down to the 140 char restriction, people choose to put the tweet out and hope people will click through to find out more. Hopefully if I start scraping the additional data it'll not turn the results upside down but just level most of them up.
Oh, I didn't mean that as a criticism. I understand why they're vague and all that — I'm just surprised you were able to get this kind of information and curious how you went about it, and thought it might be worth integrating into the front-end if it was anything fancy.
I'm planing to start tagging each tweet with keywords, which in itself may help sift some spam out. But as you pointed out a number of them are very vague which is why I've held of allowing people to filter results for fear of them missing some which may be relevant to them. I'll probably just suggest a wider result set or search for them as appropriate.
