2017/09/15 by Shereen Oraby, Oraby, Shereen, Vrindavan Harrison +7 · 1 citation
Computer Science · Psychology · #Computation and Language (cs.CL) #FOS: Computer and information sciences #Language, Metaphor, and Cognition #Sentiment Analysis and Opinion Mining #Topic Modeling
paper · pdf · doi:10.48550/arxiv.1709.05305
openalex publication_date 2017/09/15 · openalex created_date 2025/10/10 · openalex updated_date 2026/07/28
Effective models of social dialog must understand a broad range of rhetorical\nand figurative devices. Rhetorical questions (RQs) are a type of figurative\nlanguage whose aim is to achieve a pragmatic goal, such as structuring an\nargument, being persuasive, emphasizing a point, or being ironic. While there\nare computational models for other forms of figurative language, rhetorical\nquestions have received little attention to date. We expand a small dataset\nfrom previous work, presenting a corpus of 10,270 RQs from debate forums and\nTwitter that represent different discourse functions. We show that we can\nclearly distinguish between RQs and sincere questions (0.76 F1). We then show\nthat RQs can be used both sarcastically and non-sarcastically, observing that\nnon-sarcastic (other) uses of RQs are frequently argumentative in forums, and\npersuasive in tweets. We present experiments to distinguish between these uses\nof RQs using SVM and LSTM models that represent linguistic features and\npost-level context, achieving results as high as 0.76 F1 for "sarcastic" and\n0.77 F1 for "other" in forums, and 0.83 F1 for both "sarcastic" and "other" in\ntweets. We supplement our quantitative experiments with an in-depth\ncharacterization of the linguistic variation in RQs.\n