Research map / Architectures and efficiency / Attention and transformers
Explicit Sentence Compression for Neural Machine Translation
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 a compressed backbone sequence and fuse it on both source and target sides | Transformer (base) · WMT14 EN-DE newstest2014 | BLEU: 27.24 -> 28.35 | Helped |
| Add a compressed backbone sequence and fuse it on both source and target sides | Transformer (base) · WMT14 EN-FR newstest2014 | BLEU: 38.21 -> 39.40 | Helped |
| Fuse compressed backbone information on the target side rather than the source side | Transformer (base) · WMT14 EN-DE newstest2014 | BLEU: 27.75 -> 28.14 | Helped |
| Replace full-text compression with ESC-generated compressed backbone information (ablation) | BBFNMT · WMT14 EN-DE newstest2014 | BLEU: 27.24 -> 28.35 | Helped |
| Replace full-text compression with the first 8 words (ablation) | BBFNMT · WMT14 EN-DE newstest2014 | BLEU: 27.24 -> 27.40 | No clear effect |
| Replace ESC compression with random sampling (ablation) | BBFNMT · WMT14 EN-DE newstest2014 | BLEU: 27.24 -> 26.53 | Hurt |
| Add supervised ESC-generated compression (ablation) | BBFNMT · WMT14 EN-DE newstest2014 | BLEU: 27.24 -> 27.80 | Helped |
| Add unsupervised ESC-generated compression (ablation) | BBFNMT · WMT14 EN-DE newstest2014 | BLEU: 27.24 -> 27.97 | Helped |
| Fine-tune unsupervised ESC with annotated compression data (ablation) | BBFNMT · WMT14 EN-DE newstest2014 | BLEU: 27.97 -> 28.35 | Helped |
| Use independent encoders instead of a shared encoder (ablation) | BBFNMT · WMT14 EN-DE newstest2014 | BLEU: 28.35 -> 28.50 | Helped |
| Use independent encoders instead of a shared encoder (ablation) | BBFNMT · WMT14 EN-DE newstest2014 | number of parameters: 78.6M -> 91.6M | Hurt |
| Vary the compression ratio; the paper reports the best setting at gamma = 0.6 (ablation) | BBFNMT with semi-supervised ESC · WMT14 EN-DE newstest2014 | BLEU | Mixed |
Closest papers by meaning
- Neural Machine Translation with Reconstruction (AAAI 2017 · similarity 0.89)
- From Fully Trained to Fully Random Embeddings: Improving Neural Machine Translation with Compact Word Embedding Tables (AAAI 2022 · similarity 0.88)
- Deep Fusing Pre-trained Models into Neural Machine Translation (AAAI 2022 · similarity 0.88)
- Tied Transformers: Neural Machine Translation with Shared Encoder and Decoder (AAAI 2019 · similarity 0.87)
- Unsupervised Neural Machine Translation (ICLR 2018 · similarity 0.87)
- Multi-Channel Encoder for Neural Machine Translation (AAAI 2018 · 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.