[kmort]

Perhaps this is something that may be mined from the Twitter firehose (which may be easier than marketing a new site).

Let's be generous and say:

  Two mentions in the firehose would be sufficient for a correlation
  8% of people twitter
  25% of twitterers would tweet that they were food poisoned AND name the restaurant
  25% of the patrons got food poisioning

Any restaurant with at least (2 / 0.08 / 0.25 / 0.25) 400 patrons for the given time period would trip the alarm.

Those are pretty generous numbers though, with the figure jumping quite significantly as they're tweaked. A 25% poisoning rate for a reasonably-sized restaurant would probably hit the news beyond Twitter anyway.