Noise Isn’t Always Negative: Countering Exposure Bias in Sequence-to-Sequence Inflection Models

Garrett Nicolai, Miikka Silfverberg


Abstract
Morphological inflection, like many sequence-to-sequence tasks, sees great performance from recurrent neural architectures when data is plentiful, but performance falls off sharply in lower-data settings. We investigate one aspect of neural seq2seq models that we hypothesize contributes to overfitting - teacher forcing. By creating different training and test conditions, exposure bias increases the likelihood that a system too closely models its training data. Experiments show that teacher-forced models struggle to recover when they enter unknown territory. However, a simple modification to the training algorithm to more closely mimic test conditions creates models that are better able to generalize to unseen environments.
Anthology ID:
2020.coling-main.255
Volume:
Proceedings of the 28th International Conference on Computational Linguistics
Month:
December
Year:
2020
Address:
Barcelona, Spain (Online)
Editors:
Donia Scott, Nuria Bel, Chengqing Zong
Venue:
COLING
SIG:
Publisher:
International Committee on Computational Linguistics
Note:
Pages:
2837–2846
Language:
URL:
https://aclanthology.org/2020.coling-main.255
DOI:
10.18653/v1/2020.coling-main.255
Bibkey:
Cite (ACL):
Garrett Nicolai and Miikka Silfverberg. 2020. Noise Isn’t Always Negative: Countering Exposure Bias in Sequence-to-Sequence Inflection Models. In Proceedings of the 28th International Conference on Computational Linguistics, pages 2837–2846, Barcelona, Spain (Online). International Committee on Computational Linguistics.
Cite (Informal):
Noise Isn’t Always Negative: Countering Exposure Bias in Sequence-to-Sequence Inflection Models (Nicolai & Silfverberg, COLING 2020)
Copy Citation:
PDF:
https://preview.aclanthology.org/nschneid-patch-4/2020.coling-main.255.pdf