Social media posts often contain questions, but many of the questions are rhetorical and do not seek information. Our work studies the problem of distinguishing rhetorical and information-seeking questions on Twitter. Most work has focused on features of the question itself, but we hypothesize that the prior context plays a role too. This paper introduces a new dataset containing questions in tweets paired with their prior tweets to provide context. We create classification models to assess the difficulty of distinguishing rhetorical and information-seeking questions, and experiment with different properties of the prior context. Our results show that the prior tweet and topic features can improve performance on this task.

This paper presents the first research aimed at recognizing euphemistic and dysphemistic phrases with natural language processing. Euphemisms soften references to topics that are sensitive, disagreeable, or taboo. Conversely, dysphemisms refer to sensitive topics in a harsh or rude way. For example, “passed away” and “departed” are euphemisms for death, while “croaked” and “six feet under” are dysphemisms for death. Our work explores the use of sentiment analysis to recognize euphemistic and dysphemistic language. First, we identify near-synonym phrases for three topics (firing, lying, and stealing) using a bootstrapping algorithm for semantic lexicon induction. Next, we classify phrases as euphemistic, dysphemistic, or neutral using lexical sentiment cues and contextual sentiment analysis. We introduce a new gold standard data set and present our experimental results for this task.