Twitter just released the data of a few millions tweets... That should give a nice sandbox to see if a sample list would be too easy / hard.