Get Started
Home
Topics
Search
Library
Research questionHow can sign recognition from continuous signing identify unseen signs without gloss-annotated vocabularies?A fixed gloss classifier has no output for a sign outside its training inventory, while a target vocabulary may include signs without gloss labels. The system must therefore connect visual articulation in continuous-signing video with vocabulary entries despite unfamiliar signs.
AI
Computer Vision
Image & Video Processing
Information Retrieval
Machine Learning
Multimodal Models
Natural Language Processing
Research Paper
Technology
Latest papersRecent research connected to this question, newest first.A Reverse Sign Language Dictionary: Open-Vocabulary Sign Recognition from Continuous Signing via Video Captioning and Description RetrievalThe evidence comes from 1,300 sign-level segments in a Japanese Sign Language dialogue corpus, paired with procedural descriptions and a 503-entry target vocabulary. The studied system captions clips with an open-weight vision-language model and retrieves target descriptions using a multilingual sentence encoder; results cover seen- and unseen-class retrieval, with comparison to a supervised classifier only where applicable. Findings are limited to the reported JSL datasets, and the source identifies captioning quality as a remaining gap.research paper · Sep 3, 2026
Related questions
How can video-based sign dictionary retrieval generalize across signers and unseen languages?How can sign language translation models capture asynchronous lip cues and recognize fingerspelled terms without detailed supervision?How can sign-language production models generate grammatical non-manual features without losing expressive variation?How can visual speech recognition resolve ambiguous words without committing before enough context is available?