Learning Policies for Multilingual Training of Neural Machine Translation Systems
arXiv:2103.06964
Abstract
Low-resource Multilingual Neural Machine Translation (MNMT) is typically tasked with improving the translation performance on one or more language pairs with the aid of high-resource language pairs. In this paper, we propose two simple search based curricula -- orderings of the multilingual training data -- which help improve translation performance in conjunction with existing techniques such as fine-tuning. Additionally, we attempt to learn a curriculum for MNMT from scratch jointly with the training of the translation system with the aid of contextual multi-arm bandits. We show on the FLORES low-resource translation dataset that these learned curricula can provide better starting points for fine tuning and improve overall performance of the translation system.
arXiv comment7 pages, 2 figures
References in corpus (6)
- Model-Agnostic Meta-Learning for Fast Adaptation of Deep Networks
- Algorithms for multi-armed bandit problems
- Fast Domain Adaptation for Neural Machine Translation
- Bandit Problems with Side Observations
- An Empirical Exploration of Curriculum Learning for Neural Machine Translation
- Improving Zero-Shot Translation of Low-Resource Languages