[2003.02739] Zero-Shot Cross-Lingual Transfer with Meta Learningopen searchopen navigation menucontact arXivsubscribe to arXiv mailings

Learning what to share between tasks has been a topic of great importance recently, as strategic sharing of knowledge has been shown to improve downstream task performance. This is particularly important for multilingual applications, as most languages in the world are under-resourced. Here, we consider the setting of training models on multiple different languages at the same time, when little or no data is available for languages other than English. We show that this challenging setup can be approached using meta-learning, where, in addition to training a source language model, another model learns to select which training instances are the most beneficial to the first. We experiment using standard supervised, zero-shot cross-lingual, as well as few-shot cross-lingual settings for different natural language understanding tasks (natural language inference, question answering). Our extensive experimental setup demonstrates the consistent effectiveness of meta-learning for a total of 15

3 mentions: @farhadnz@CopeNLU@johannesbjerva
Date: 2020/09/15 11:22

Referring Tweets

@farhadnz I’m glad to share that our paper "Zero-Shot Cross-Lingual Transfer with Meta Learning" t.co/62pPwgC2gp is accepted to the main conference of @emnlp2020! I’d like to thank my wonderful co-authors @g_bekou, @johannesbjerva and @IAugenstein . @CopeNLU @ltgoslo
@CopeNLU "Zero-Shot Cross-Lingual Transfer with Meta Learning" (@farhadnz @g_bekou @johannesbjerva @IAugenstein) shows how meta-learning with one or several auxiliary languages can improve zero- or few-shot NLU performance. Preprint: t.co/OqL22pOxFa #emnlp2020 t.co/69NyVK0upC

Related Entries

Read more GitHub - Zhongdao/Towards-Realtime-MOT: Joint Detection and Embedding for fast multi-object tracking
1 users, 12 mentions 2019/10/07 15:48
Read more GitHub - mfinzi/LieConv
0 users, 2 mentions 2020/03/02 02:21
Read more Robotics Retrospectives
0 users, 1 mentions 2020/03/11 18:51
Read more GitHub - iduta/iresnet: Improved Residual Networks
0 users, 5 mentions 2020/04/14 20:20
Read more Concept2Robot: Learning Manipulation Concepts from Instructions and Human Demonstrations · Robotics:...
0 users, 1 mentions 2020/07/16 05:21