Research map / Label-efficient and robust learning / Recognition and long-tail
Addressing the Under-Translation Problem from the Entropy Perspective
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 two-pass pseudo-target method | RNMT (word) · NIST CH-EN avg | BLEU: 39.16 -> 40.37 | Helped |
| add multitask pseudo-target method | RNMT (sub-word) · NIST CH-EN avg | BLEU: 41.34 -> 41.98 | Helped |
| add pre-training on pseudo targets | RNMT (word) · NIST CH-EN avg | BLEU: 39.16 -> 39.58 | Helped |
| add two-pass method on top of coverage model | RNMT+coverage · NIST CH-EN avg | BLEU: 39.94 -> 40.87 | Helped |
| add two-pass method | Transformer · NIST CH-EN avg | BLEU: 44.40 -> 45.04 | Helped |
| add two-pass method | RNMT · KFTT EN-JA | BLEU: 27.43 -> 28.56 | Helped |
| add two-pass method | Transformer · WMT14 EN-DE | BLEU: 27.20 -> 27.73 | Helped |
| add two-pass method | RNMT · CH-EN high-entropy words | under-translation ratio: 8.05% -> 4.86% | Helped |
| add two-pass method | RNMT · CH-EN other words | under-translation ratio: 3.88% -> 3.81% | No clear effect |
| add two-pass method | Transformer · CH-EN high-entropy words | AER: 49.47 -> 44.02 | Helped |
| lower entropy threshold e0 from 8 to 4 (ablation) | RNMT+two pass · NIST CH-EN avg | BLEU: 39.72 -> 40.37 | Helped |
| lower entropy threshold e0 from 4 to 2 (ablation) | RNMT+two pass · NIST CH-EN avg | BLEU: 40.37 -> 40.30 | No clear effect |
Closest papers by meaning
- Scaling Laws for Neural Machine Translation (ICLR 2022 · similarity 0.87)
- Unsupervised Neural Machine Translation (ICLR 2018 · similarity 0.87)
- Unsupervised Neural Machine Translation with SMT as Posterior Regularization (AAAI 2019 · similarity 0.86)
- Mixed Cross Entropy Loss for Neural Machine Translation (ICML 2021 · similarity 0.86)
- Improved Neural Machine Translation with SMT Features (AAAI 2016 · similarity 0.86)
- A Paradigm Shift in Machine Translation: Boosting Translation Performance of Large Language Models (ICLR 2024 · similarity 0.84)
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.