Unverified50% confidenceFactExact time
Academia typically decomposes sign language translation into two steps: sign language recognition converts video into a sign language word sequence, which is then converted into target language text; end-to-end approaches are still behind cascaded solutions in performance due to the scarcity of parallel corpora.
1
Sources
50%
Confidence
Long-term
Relevance
8/4/2026
First Seen
Sources
Hand Wave：Meta智能眼镜实时手语翻译，让听障沟通无障碍
Related Claims
UnverifiedThe input for sign language translation is a continuous video stream requiring simultaneous completion of three levels of tasks—visual perception, temporal segmentation, and semantic understanding—making it more complex than text-to-text machine translation.78% similarUnverified手语翻译的端到端方法受限于平行语料稀缺性，目前性能仍落后于级联方案74% similarUnverified语码转换分为句间转换（inter-sentential）和句内转换（intra-sentential），技术从业者口述通常属于句内转换70% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/684587API
curl https://kongchang.com/api/v1/knowledge/claims/684587MCP
get_claim(id=684587)