Trajectory-Based Meta-Learning for Out-Of-Vocabulary Word Embedding Learning

Gordon Buck, Andreas Vlachos


Abstract
Word embedding learning methods require a large number of occurrences of a word to accurately learn its embedding. However, out-of-vocabulary (OOV) words which do not appear in the training corpus emerge frequently in the smaller downstream data. Recent work formulated OOV embedding learning as a few-shot regression problem and demonstrated that meta-learning can improve results obtained. However, the algorithm used, model-agnostic meta-learning (MAML) is known to be unstable and perform worse when a large number of gradient steps are used for parameter updates. In this work, we propose the use of Leap, a meta-learning algorithm which leverages the entire trajectory of the learning process instead of just the beginning and the end points, and thus ameliorates these two issues. In our experiments on a benchmark OOV embedding learning dataset and in an extrinsic evaluation, Leap performs comparably or better than MAML. We go on to examine which contexts are most beneficial to learn an OOV embedding from, and propose that the choice of contexts may matter more than the meta-learning employed.
Anthology ID:
2021.adaptnlp-1.15
Volume:
Proceedings of the Second Workshop on Domain Adaptation for NLP
Month:
April
Year:
2021
Address:
Kyiv, Ukraine
Editors:
Eyal Ben-David, Shay Cohen, Ryan McDonald, Barbara Plank, Roi Reichart, Guy Rotman, Yftah Ziser
Venue:
AdaptNLP
SIG:
Publisher:
Association for Computational Linguistics
Note:
Pages:
146–155
Language:
URL:
https://aclanthology.org/2021.adaptnlp-1.15
DOI:
Bibkey:
Cite (ACL):
Gordon Buck and Andreas Vlachos. 2021. Trajectory-Based Meta-Learning for Out-Of-Vocabulary Word Embedding Learning. In Proceedings of the Second Workshop on Domain Adaptation for NLP, pages 146–155, Kyiv, Ukraine. Association for Computational Linguistics.
Cite (Informal):
Trajectory-Based Meta-Learning for Out-Of-Vocabulary Word Embedding Learning (Buck & Vlachos, AdaptNLP 2021)
Copy Citation:
PDF:
https://aclanthology.org/2021.adaptnlp-1.15.pdf
Data
WikiText-103WikiText-2