Despite being a complex theoretical concept, sarcasm or verbal irony detection has become one of the popular tasks in NLP in the last decade. In this talk, rather than discussing the state of the arts in sarcasm detection, I will focus on some of the new directions– interpretation, generation, and novel applications – of sarcasm. First, I will explain how humans interpret sarcasm and the ties between such interpretations with NLP tasks, such as RTE. Next, I will discuss how common sense is applicable to generate sarcasm in a retrieve and rank framework. Finally, I will point out the role of sarcasm in shaping up the dis(agreement) space in an argumentative discussion forum.
Debanjan Ghosh is a NLP Research Scientist in the AI Lab at Educational Testing Service (ETS) in Princeton, NJ. Before joining ETS, Debanjan was a postdoctoral researcher at MIT, in Brain and Cognitive Science Department. Prior to that, he obtained his PhD from Rutgers University. His main areas of research are computational linguistics and artificial intelligence, particularly in figurative language, argument mining, and natural language generation. At ETS, his current focus is on question generation for Reading Comprehension as well as dialog generations for educational applications. Debanjan and his collaborators are the recipients of the best paper award at SIGDial in 2017, and the 3rd prize at the NYC Media Lab Summit 2015 for their research on Sarcasm Detection.