In training set, the last element in the list is the label.
In validation and testing set, the last but one element in the list is labeled by hashtag(original post). The last element in the list is labeled by another human annotator.
The number of posts is bigger than that are used in the code. Since in the code, posts with these words (exgag, sarcasm, sarcastic, reposting, , joke, humour, humor, jokes, irony, ironic) are discarded. Images of these posts are not uploaded.