We research the use of sarcasm on Twitter, and show that a computer has more difficulty to detect sarcasm shared among peers than sarcasm shared with any interested audience. This data set features the data used for training machine learning classifiers, and annotations of the output.
