Mechanical Turk is a service that allows us to create and manage online human intelligence tasks that will be completed by human workers. There are lots of tasks that humans can do better than computers. Therefore, we can take advantage of this service to support our machine learning system.

You can view this system at https://www.mturk.com. Here is the website of the service:

Here are a couple of examples of tasks that you can use to support your machine learning system:

• Dataset labeling: You usually have a lot of unlabeled data, and you can use Mechanical Turk to help you build a consistent ground truth for your machine learning workflow.
• Generate dataset: You can ask the workers to build a large amount of training data. For example, we can ask workers to create text translations or chat sentences for a natural language system. You can ask them to annotate the sentiments of the comments.

Beyond labeling, Mechanical Turk can also clean up your messy datasets ready for training, data categorization, and metadata tagging. You can even use this service to have them judge your system output.