Task Description
This is a text classification task. Every document (a line in the data file) contains a Reddit comment and its parent post (the post being commented on), separated with tab.
Your goal is to classify each Reddit comment (not the parent post) into ONE of the two categories, based on whether it is sarcastic or not. The parent post provides necessary context for your judgment.
**0: the comment does NOT express sarcasm**

**1: the comment expresses sarcasm**
The training data contains 53,032 comments, already labeled with one of the above two categories. The test data contains 17,719 comments that are unlabeled. The submission should be a .csv (comma separated free text) file with a header line "id,label" followed by exactly 17,719 lines. In each line, there should be exactly two integers, separated by a comma. The first integer is the line ID of a test example (0, 1, 2, ..., 17718), and the second integer is the category (or label) predicted by your classifier, one of {0,1}.
You can make 10 submissions per day. Once you submit your results, you will get an accuracy score computed based on 50% of the test data. This score will position you somewhere on the leaderboard. Once the competition ends, you will see the final accuracy computed based on the other 50% of the test data. The evaluation metric is the accuracy of your classifier. It is calculated as (# of correctly classified test examples)/(total # of test examples), so the higher the better.
You can use any classifier or combination of classifiers, any combination or selection of features, and either supervised, semi-supervised, or even transfer learning approaches. You can also be creative and make use of external data sources that do not contain the exact comments or parent posts in Reddit, or the original larger data set by Dan Ofer.
Acknowledgements
We thank Dan Ofer and Professor Plum for providing the data.