RiddleSense: Reasoning about Riddle Questions Featuring Linguistic Creativity and Commonsense Knowledge

TL;DR

RiddleSense introduces a new dataset and task for evaluating advanced natural language understanding, focusing on reasoning about riddles that require commonsense, figurative language, and counterfactual thinking.

Contribution

The paper presents RiddleSense, the first large dataset of 5.7k riddles for testing complex reasoning in NLU, and systematically evaluates models highlighting significant gaps to human performance.

Findings

Models lag behind humans in solving riddles.

The dataset reveals challenges in commonsense and figurative language understanding.

Current models show limited reasoning capabilities on the task.

Abstract

Question: I have five fingers but I am not alive. What am I? Answer: a glove. Answering such a riddle-style question is a challenging cognitive process, in that it requires complex commonsense reasoning abilities, an understanding of figurative language, and counterfactual reasoning skills, which are all important abilities for advanced natural language understanding (NLU). However, there are currently no dedicated datasets aiming to test these abilities. Herein, we present RiddleSense, a new multiple-choice question answering task, which comes with the first large dataset (5.7k examples) for answering riddle-style commonsense questions. We systematically evaluate a wide range of models over the challenge, and point out that there is a large gap between the best-supervised model and human performance -- suggesting intriguing future research in the direction of higher-order commonsense reasoning and linguistic creativity towards building advanced NLU systems.