CICERO: A Dataset for Contextualized Commonsense Inference in Dialogues

This paper addresses the problem of dialogue reasoning with contextualized commonsense inference. We curate CICERO, a dataset of dyadic conversations with five types of utterance-level reasoning-based inferences: cause, subsequent event, prerequisite, motivation, and emotional reaction. The dataset contains 53,105 of such inferences from 5,672 dialogues. We use this dataset to solve relevant generative and discriminative tasks: generation of cause and subsequent event; generation of prerequisite, motivation, and listener's emotional reaction; and selection of plausible alternatives. Our results ascertain the value of such dialogue-centric commonsense knowledge datasets. It is our hope that CI-CERO will open new research avenues into commonsense-based dialogue reasoning. A: Can I help you? B: Yes, please. I'd like some oranges. A: Do you want Florida or California oranges? B: Which do you think are better? A: Florida oranges are sweet but they are small. But California oranges have no seeds. B: Then give me five California oranges. A: Anything else? B: I also want some bananas. How do you sell them? A: One dollar a pound. How many do you want? B: Give me four and see how much they are.