Research map / Large language models / Language, safety and interpretability
Improving Sign Language Translation With Monolingual Data by Sign Back-Translation
Check 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 SignBT-generated training pairs | BN-TIN-Transf. S2T · CSL-Daily test | BLEU-4: 13.19 -> 21.34 | Helped |
| Add SignBT-generated training pairs | BN-TIN-Transf.² S2G2T · CSL-Daily test | BLEU-4: 16.25 -> 19.67 | Helped |
| Add SignBT-generated training pairs | BN-TIN-Transf. S2T · PHOENIX-2014T test | BLEU-4: 21.68 -> 24.32 | Helped |
| Remove SignBT-generated training pairs (ablation) | Transformer S2G2T · CSL-Daily Dev | BLEU-4: 19.53 -> 15.93 | Hurt |
| Increase synthetic data quantity to >30× parallel data (ablation) | Transformer S2T · CSL-Daily Dev | BLEU-4: 12.73 -> 20.80 | Helped |
| Add synthetic-data warm-up (ablation) | Transformer S2G2T · CSL-Daily Dev | BLEU-4: 15.93 -> 19.53 | Helped |
| Use 100 rather than 50 warm-up epochs (ablation) | Transformer S2G2T · CSL-Daily Dev | BLEU-4: 19.53 -> 19.42 | Hurt |
| Use 100 rather than 50 warm-up epochs (ablation) | Transformer S2T · CSL-Daily Dev | BLEU-4: 20.20 -> 20.50 | Helped |
| Add a small synthetic-data share after warm-up (ablation) | Transformer S2T · CSL-Daily Dev | BLEU-4: 20.20 -> 20.80 | Helped |
| Add synthetic data after warm-up (ablation) | Transformer S2G2T · CSL-Daily Dev | BLEU-4: 19.53 -> 19.20 | Hurt |
| Improve text-to-gloss synthetic-pair quality (ablation) | Transformer S2T · CSL-Daily Dev | BLEU-4: 15.77 -> 20.80 | Helped |
| Replace synthetic sign features with blank input (ablation) | Transformer S2T · CSL-Daily Dev | BLEU-4: 20.80 -> 15.77 | Hurt |
Closest papers by meaning
- Scaling Sign Language Translation (NeurIPS 2024 · similarity 0.93)
- Sign2GPT: Leveraging Large Language Models for Gloss-Free Sign Language Translation (ICLR 2024 · similarity 0.92)
- Bridging Sign and Spoken Languages: Pseudo Gloss Generation for Sign Language Translation (NeurIPS 2025 · similarity 0.91)
- LLMs are Good Sign Language Translators (CVPR 2024 · similarity 0.91)
- Improving Gloss-free Sign Language Translation by Reducing Representation Density (NeurIPS 2024 · similarity 0.90)
- Lost in Translation, Found in Context: Sign Language Translation with Contextual Cues (CVPR 2025 · similarity 0.90)
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.