Synapse
⌘+K
Synapse
PulseExploreClubsResearchersJournals
Instagram
HomeClubsExplore
August 5, 2026ACM Transactions on Asian and Low-Resource Language Information Processing

Enhancing Myanmar Sign Language Translation Through Transfer Learning, Self-training, and Error Correction

View Full Paper
Ask AI
Bookmark
Share

Authors

HNHlaing Myat NweKSKiyoaki ShiraiNKNatthawut Kertkeidkachorn

Discussion

Loading...

Member takes

Overview

Randomized trial investigates machine translation improvements for Myanmar Sign Language, suggesting enhanced accuracy through innovative methods.

Key Points

  • The aim is to improve machine translation between Myanmar Sign Language and Myanmar Written Language using advanced techniques.
  • Implemented fine-tuning of multilingual models mT5 and mBART-50.
  • Applied transfer learning using a large-scale parallel corpus of American Sign Language and English.
  • Conducted error correction by fine-tuning with synthetic noisy data.
  • The hybrid approach significantly improved translation quality between MSL and MWL.
  • Transfer learning and self-training both contributed positively to MT performance.
  • Error correction enhanced grammatical accuracy in translated sentences.

Cite This Study

Nwe et al. (2026) studied this question.

synapsesocial.com/papers/6a72e7f4226790f37065762bhttps://doi.org/10.1145/3837853
View Full Paper
Ask AI
Bookmark
Share