Jump to better conclusions: SCAN both left and right

Jasmijn Bastings, Marco Baroni, Jason Weston, Kyunghyun Cho, Douwe Kiela


Abstract
Lake and Baroni (2018) recently introduced the SCAN data set, which consists of simple commands paired with action sequences and is intended to test the strong generalization abilities of recurrent sequence-to-sequence models. Their initial experiments suggested that such models may fail because they lack the ability to extract systematic rules. Here, we take a closer look at SCAN and show that it does not always capture the kind of generalization that it was designed for. To mitigate this we propose a complementary dataset, which requires mapping actions back to the original commands, called NACS. We show that models that do well on SCAN do not necessarily do well on NACS, and that NACS exhibits properties more closely aligned with realistic use-cases for sequence-to-sequence models.
Anthology ID:
W18-5407
Original:
W18-5407v1
Version 2:
W18-5407v2
Volume:
Proceedings of the 2018 EMNLP Workshop BlackboxNLP: Analyzing and Interpreting Neural Networks for NLP
Month:
November
Year:
2018
Address:
Brussels, Belgium
Venue:
EMNLP
SIG:
Publisher:
Association for Computational Linguistics
Note:
Pages:
47–55
Language:
URL:
https://aclanthology.org/W18-5407
DOI:
10.18653/v1/W18-5407
Bibkey:
Cite (ACL):
Jasmijn Bastings, Marco Baroni, Jason Weston, Kyunghyun Cho, and Douwe Kiela. 2018. Jump to better conclusions: SCAN both left and right. In Proceedings of the 2018 EMNLP Workshop BlackboxNLP: Analyzing and Interpreting Neural Networks for NLP, pages 47–55, Brussels, Belgium. Association for Computational Linguistics.
Cite (Informal):
Jump to better conclusions: SCAN both left and right (Bastings et al., EMNLP 2018)
Copy Citation:
PDF:
https://preview.aclanthology.org/ingestion-script-update/W18-5407.pdf
Code
 facebookresearch/NACS
Data
SCAN