Research map / Architectures and efficiency / RNNs and normalization
DTMT: A Novel Deep Transition Architecture for Neural Machine Translation
Fandong Meng, Jinchao Zhang
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (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 |
|---|---|---|---|
| Increase transition depth from the shallow baseline to DTMT#4 | RNMT · NIST Zh⇒En average test sets | BLEU: 41.57 -> 45.04 | Improved |
| Increase transition depth from DTMT#1 to DTMT#4 (ablation) | DTMT · NIST Zh⇒En average test sets | BLEU: 44.41 -> 45.04 | Improved |
| Replace GRU with L-GRU (ablation) | ShallowRNMT · NIST Zh⇒En average test sets | BLEU: 41.57 -> 43.83 | Improved |
| Replace LAU with L-GRU (ablation) | ShallowRNMT · NIST Zh⇒En average test sets | BLEU: 43.06 -> 43.83 | Improved |
| Replace GRU with L-GRU (ablation) | DTMT#1 · NIST Zh⇒En average test sets | BLEU: 43.63 -> 44.41 | Improved |
| Replace LAU with L-GRU (ablation) | DTMT#1 · NIST Zh⇒En average test sets | BLEU: 43.79 -> 44.41 | Improved |
| Replace GRU with L-GRU (ablation) | DTMT#4 · NIST Zh⇒En average test sets | BLEU: 44.16 -> 45.04 | Improved |
| Replace LAU with L-GRU (ablation) | DTMT#4 · NIST Zh⇒En average test sets | BLEU: 44.54 -> 45.04 | Improved |
| Add encoder transition rather than using the shallow RNMT encoder component (ablation) | DTMT#4 · NIST Zh⇒En average test sets | BLEU: 43.30 -> 45.04 | Improved |
| Add query transition rather than using the shallow RNMT component (ablation) | DTMT#4 · NIST Zh⇒En average test sets | BLEU: 44.35 -> 45.04 | Improved |
| Add decoder transition rather than using the shallow RNMT component (ablation) | DTMT#4 · NIST Zh⇒En average test sets | BLEU: 44.36 -> 45.04 | Improved |
| Add L-GRU (ablation) | DTMT#4 · WMT14 En⇒De | BLEU: 27.81 -> 28.70 | Improved |
Closest papers by meaning
- Deep Fusing Pre-trained Models into Neural Machine Translation (AAAI 2022 · similarity 0.87)
- Asynchronous Bidirectional Decoding for Neural Machine Translation (AAAI 2018 · similarity 0.87)
- Towards Making the Most of BERT in Neural Machine Translation (AAAI 2020 · similarity 0.87)
- Layer-Wise Coordination between Encoder and Decoder for Neural Machine Translation (NeurIPS 2018 · similarity 0.86)
- Variational Recurrent Neural Machine Translation (AAAI 2018 · similarity 0.86)
- Improved Neural Machine Translation with SMT Features (AAAI 2016 · similarity 0.86)
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.
