The MSLG-SPA 2026 shared task focuses on the automatic translation between Mexican Sign Language glosses (MSL-G) and Spanish in a bidirectional setting. Unlike spoken language translation, sign languages present unique challenges due to their visual-gestural modality, rich morphology, and the lack of standardized written forms. Glosses serve as an intermediate symbolic representation that enables computational processing and evaluation.
The task is divided into two complementary subtasks:
Subtask A: MSL Gloss → Spanish (MSLG2SPA)
Participants develop systems that take sequences of MSL glosses as input and generate fluent and semantically accurate Spanish sentences.
Subtask B: Spanish → MSL Gloss (SPA2MSLG)
Participants develop systems that translate Spanish sentences into appropriate sequences of MSL glosses, capturing the semantic content and grammatical structure of MSL.
Establish the first benchmark for MSL gloss–Spanish translation.
Provide a common evaluation protocol for comparing models in both directions.
Encourage the development of data-driven and linguistically informed approaches to sign language processing.
Support future research toward multimodal translation systems involving video, glosses, and spoken language.