There has been a recent work in the same domain where neural networks(CNNs to be accurate) are used for the same purpose. Some info. about the research is:
To learn that context, the paper describes a method by which the
neural network finds the user’s “embeddings” — i.e. contextual cues
like the content of previous tweets, related interests and accounts,
and so on. It uses these various factors to plot the user with others,
and (ideally) finds that they form relatively well-defined groups.
So, the paper uses CNNs, word and user embeddings for detecting sarcasm in text. There is also a Techcrunch article on that.
The paper uses sentiment of the tweet and compares with that of the other similar tweets:
If the sentiment of the tweet seems to disagree with the bulk of what
is expressed by similar users, there’s a good chance sarcasm is being
Link to the paper