Research map / Label-efficient and robust learning / Domain adaptation and generalization
Dual Transfer Learning for Neural Machine Translation with Marginal Distribution Regularization
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add marginal distribution regularization with monolingual data | RNNSearch · WMT14 En-Fr | BLEU: 29.92 -> 32.85 | Helped |
| add marginal distribution regularization with monolingual data | 2-layer LSTM · IWSLT14 De-En | BLEU: 30.99 -> 32.35 | Helped |
| add marginal distribution regularization with monolingual data | 4-4 layer LSTM + BPE · WMT14 En-Fr | BLEU: 38.80 -> 39.98 | Helped |
| dual transfer learning instead of Dual-NMT | RNNSearch · WMT14 En-Fr | BLEU: 32.06 -> 32.85 | Helped |
| dual transfer learning instead of Dual-NMT | 2-layer LSTM · IWSLT14 De-En | BLEU: 32.05 -> 32.35 | Helped |
| shallow fusion LM instead of no monolingual data | RNNSearch · WMT14 En-Fr | BLEU: 29.92 -> 30.03 | No clear effect |
| increase unlabeled data ratio up to 0.8 (ablation) | dual transfer learning · IWSLT14 De-En valid | BLEU | Helped |
| increase unlabeled data ratio beyond 0.8 (ablation) | dual transfer learning · IWSLT14 De-En valid | BLEU | No clear effect |
| vary lambda away from 0.05 (ablation) | dual transfer learning · IWSLT14 De-En valid | BLEU | Hurt |
| increase sample size K (ablation) | dual transfer learning · IWSLT14 De-En valid | BLEU | Mixed |
| use better dual model for sampling (BLEU 17.30 to 23.94) (ablation) | dual transfer learning · IWSLT14 De-En valid | BLEU | Helped |
Closest papers by meaning
- Dual Learning for Machine Translation (NeurIPS 2016 · similarity 0.92)
- Joint Training for Neural Machine Translation Models with Monolingual Data (AAAI 2018 · similarity 0.92)
- Unsupervised Neural Machine Translation (ICLR 2018 · similarity 0.90)
- Unsupervised Machine Translation Using Monolingual Corpora Only (ICLR 2018 · similarity 0.89)
- Improving Simultaneous Machine Translation with Monolingual Data (AAAI 2023 · similarity 0.89)
- A Paradigm Shift in Machine Translation: Boosting Translation Performance of Large Language Models (ICLR 2024 · similarity 0.87)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.