JECC: Commonsense Reasoning Tasks Derived from Interactive Fictions

Mo Yu, Yi Gu, Xiaoxiao Guo, Yufei Feng, Xiaodan Zhu, Michael Greenspan, Murray Campbell, Chuang Gan


Abstract
Commonsense reasoning simulates the human ability to make presumptions about our physical world, and it is an essential cornerstone in building general AI systems. We proposea new commonsense reasoning dataset based on human’s Interactive Fiction (IF) gameplaywalkthroughs as human players demonstrate plentiful and diverse commonsense reasoning. The new dataset provides a natural mixture of various reasoning types and requires multi-hopreasoning. Moreover, the IF game-based construction procedure requires much less humaninterventions than previous ones. Different from existing benchmarks, our dataset focuseson the assessment of functional commonsense knowledge rules rather than factual knowledge. Hence, in order to achieve higher performance on our tasks, models need to effectively uti-lize such functional knowledge to infer the outcomes of actions, rather than relying solely onmemorizing facts. Experiments show that the introduced dataset is challenging to previousmachine reading models as well as the new large language models with a significant 20%performance gap compared to human experts.
Anthology ID:
2023.findings-acl.713
Volume:
Findings of the Association for Computational Linguistics: ACL 2023
Month:
July
Year:
2023
Address:
Toronto, Canada
Editors:
Anna Rogers, Jordan Boyd-Graber, Naoaki Okazaki
Venue:
Findings
SIG:
Publisher:
Association for Computational Linguistics
Note:
Pages:
11226–11238
Language:
URL:
https://aclanthology.org/2023.findings-acl.713
DOI:
10.18653/v1/2023.findings-acl.713
Bibkey:
Cite (ACL):
Mo Yu, Yi Gu, Xiaoxiao Guo, Yufei Feng, Xiaodan Zhu, Michael Greenspan, Murray Campbell, and Chuang Gan. 2023. JECC: Commonsense Reasoning Tasks Derived from Interactive Fictions. In Findings of the Association for Computational Linguistics: ACL 2023, pages 11226–11238, Toronto, Canada. Association for Computational Linguistics.
Cite (Informal):
JECC: Commonsense Reasoning Tasks Derived from Interactive Fictions (Yu et al., Findings 2023)
Copy Citation:
PDF:
https://aclanthology.org/2023.findings-acl.713.pdf