I am looking for services ML researchers (or hobbyists) use to get training data for their models.<p>If you use one of the popular services like Amazon's Mechanical Turk, or Crowdflower, what kind of tasks do you submit to those services.<p>Are tasks typically a classification tasks, where user is presented with a set of classes and is required to select one (or more) classes that identify the task at hand.<p>I am collecting some data, as part of a project. Please be kind, and let me know if you any of what I asked doesn't make any sense.