CLIP Colloquium: Amr Sharaf (CS)

Meta-Learning for Few-Shot NMT Adaptation

Abstract: In this talk I'll present Meta-MT, a meta-learning approach to adapt Neural Machine Translation (NMT) systems in a few-shot setting. Meta-MT provides a new approach to make NMT models  easily adaptable to many target  domains with the minimal  amount of in-domain data. We frame the adaptation of NMT systems as a meta-learning problem, where we learn to adapt to new unseen domains based on simulated offline meta-training domain adaptation tasks. We evaluate the proposed meta-learning strategy on ten domains with general large scale NMT systems. We show that Meta-MT significantly outperforms classical domain adaptation when very few in-domain examples are available. Our experiments shows that Meta-MT can outperform classical fine-tuning by up to 2.5 BLEU points after seeing only 4,000 translated words (300 parallel sentences).

Bio: Amr Sharaf is a PhD candidate in the Computational Linguistics and Information Processing (CLIP) Lab at the University of Maryland, advised by Hal Daumé III. His research focuses on developing meta-learning algorithms in the context of structured prediction for AI and NLP. He is interested in applying reinforcement and imitation learning algorithms for meta-learning and structured prediction problems in weakly supervised settings.