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 CICERO will open new research avenues into commonsense-based dialogue reasoning.

PDF Abstract ACL 2022 PDF ACL 2022 Abstract

Datasets


Introduced in the Paper:

CICERO

Used in the Paper:

DailyDialog DREAM MuTual

Results from the Paper


Task Dataset Model Metric Name Metric Value Global Rank Result Benchmark
Answer Generation CICERO T5-large pre-trained on GLUCOSE ROUGE 0.2980 # 1
Answer Generation CICERO T5-large ROUGE 0.2947 # 2
Generative Question Answering CICERO T5-large pre-trained on COMET ROUGE 0.2878 # 3
Generative Question Answering CICERO BART ROUGE 0.2837 # 4
Answer Selection CICERO Unified QA Exact Match 77.51 # 2
Generative Question Answering CICERO T5-large pre-trained on GLUCOSE ROUGE 0.2980 # 1
Generative Question Answering CICERO T5-large ROUGE 0.2946 # 2
Answer Selection CICERO T5-large Exact Match 77.68 # 1

Methods


No methods listed for this paper. Add relevant methods here