Research map / Label-efficient and robust learning / Recognition and long-tail
Joint Training for Neural Machine Translation Models with Monolingual Data
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 |
|---|---|---|---|
| Jointly train both translation directions with source- and target-side monolingual data | RNNSearch · Chinese→English NIST test sets (Average) | BLEU: 34.97 -> 39.67 | Helped |
| Jointly train both translation directions with source- and target-side monolingual data | RNNSearch · English→Chinese NIST test sets (Average) | BLEU: 15.84 -> 20.30 | Helped |
| Jointly train both translation directions with source- and target-side monolingual data | RNNSearch · English→German news-test2014 | BLEU: 19.78 -> 23.60 | Helped |
| Jointly train both translation directions with source- and target-side monolingual data | RNNSearch · German→English news-test2014 | BLEU: 24.91 -> 27.98 | Helped |
| Add joint bidirectional training rather than using monolingual back-translation alone | RNNSearch+M · Chinese→English NIST test sets (Average) | BLEU: 37.83 -> 39.67 | Helped |
| Add joint bidirectional training rather than using monolingual back-translation alone | RNNSearch+M · English→Chinese NIST test sets (Average) | BLEU: 18.87 -> 20.30 | Helped |
| Add joint bidirectional training rather than using monolingual back-translation alone | RNNSearch+M · English→German news-test2014 | BLEU: 21.89 -> 23.60 | Helped |
| Add joint bidirectional training rather than using monolingual back-translation alone | RNNSearch+M · German→English news-test2014 | BLEU: 26.81 -> 27.98 | Helped |
| Add translation-probability weighting to synthetic sentence pairs (ablation) | JT-NMT · Chinese→English NIST test sets (Average) | BLEU: 37.83 -> 38.23 | Helped |
| Add translation-probability weighting to synthetic sentence pairs (ablation) | JT-NMT · English→Chinese NIST test sets (Average) | BLEU: 18.87 -> 19.10 | Helped |
| Add translation-probability weighting to synthetic sentence pairs (ablation) | JT-NMT · German→English news-test2014 | BLEU: 26.81 -> 27.07 | Helped |
| Add translation-probability weighting to synthetic sentence pairs (ablation) | JT-NMT · English→German news-test2014 | BLEU: 21.89 -> 22.20 | Helped |
Closest papers by meaning
- Dual Transfer Learning for Neural Machine Translation with Marginal Distribution Regularization (AAAI 2018 · similarity 0.92)
- Improving Simultaneous Machine Translation with Monolingual Data (AAAI 2023 · similarity 0.91)
- Unsupervised Neural Machine Translation (ICLR 2018 · similarity 0.90)
- Neural Machine Translation Advised by Statistical Machine Translation (AAAI 2017 · similarity 0.90)
- Unsupervised Neural Machine Translation with SMT as Posterior Regularization (AAAI 2019 · similarity 0.90)
- 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.