Research map / Large language models / Language, safety and interpretability
Improving Continuous Sign Language Recognition with Cross-Lingual Signs
Fangyun Wei, Yutong Chen
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 |
|---|---|---|---|
| add CSL-Daily as cross-lingual auxiliary data with mapped labels | TwoStream-SLR · Phoenix-2014T test | WER: 19.3 -> 18.6 | Improved |
| add Phoenix-2014 as a second auxiliary dataset besides CSL-Daily | CSL-Daily->Phoenix-2014T · Phoenix-2014T dev | WER: 17.3 -> 16.9 | Improved |
| add CSL-Daily as cross-lingual auxiliary data with mapped labels | TwoStream-SLR · Phoenix-2014 test | WER: 18.8 -> 18.5 | Improved |
| add Phoenix-2014T as a second auxiliary dataset besides CSL-Daily | CSL-Daily->Phoenix-2014 · Phoenix-2014 test | WER: 18.5 -> 16.7 | Improved |
| apply cross-lingual sign mapping instead of plain multi-task learning with original CSL labels (ablation) | SingleStream-SLR · Phoenix-2014T test | WER: 21.9 -> 21.3 | Improved |
| multi-task learning with unmapped CSL-Daily over monolingual baseline (ablation) | SingleStream-SLR · Phoenix-2014T test | WER: 22.4 -> 21.9 | Improved |
| use class-level instead of instance-level mapping (ablation) | cross-lingual prediction mapping · Phoenix-2014T test | WER: 21.5 -> 21.3 | Improved |
| use cross-lingual prediction instead of dot-product of weight matrices (ablation) | class-level mapping · Phoenix-2014T test | WER: 21.9 -> 21.3 | Improved |
| raise mapping confidence threshold (map fewer auxiliary signs) (ablation) | CSL-Daily->Phoenix-2014T · Phoenix-2014T | WER | No clear effect |
| vary auxiliary sampling ratio alpha (ablation) | CSL-Daily->Phoenix-2014T · Phoenix-2014T test | WER: 22.4 -> 21.3 (alpha 0 -> 0.2) | Mixed |
| use Phoenix (DGS) as auxiliary for full CSL-Daily as primary | CSL-Daily · CSL-Daily full | WER | No clear effect |
| use Phoenix-2014T as auxiliary for scarce CSL-Daily subset | CSL-Daily 20% subset · CSL-Daily 20% subset | WER: 45.2 -> 44.0 (dev) | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- C2SLR — C2SLR: Consistency-Enhanced Continuous Sign Language Recognition (CVPR 2022) · better on 1 of 1 shared result
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Cross-Sentence Gloss Consistency for Continuous Sign Language Recognition (AAAI 2024 · similarity 0.91)
- Lost in Translation, Found in Context: Sign Language Translation with Contextual Cues (CVPR 2025 · similarity 0.89)
- TCNet: Continuous Sign Language Recognition from Trajectories and Correlated Regions (AAAI 2024 · similarity 0.89)
- MM-WLAuslan: Multi-View Multi-Modal Word-Level Australian Sign Language Recognition Dataset (NeurIPS 2024 · similarity 0.88)
- KD-MSLRT: Lightweight Sign Language Recognition Model Based on Mediapipe and 3D to 1D Knowledge Distillation (AAAI 2025 · similarity 0.88)
- Uni-Sign: Toward Unified Sign Language Understanding at Scale (ICLR 2025 · similarity 0.88)
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.
