HATL: Hierarchical Adaptive-Transfer Learning Framework for Sign Language Machine Translation

TL;DR

This paper introduces HATL, a hierarchical adaptive transfer learning framework for sign language translation that dynamically unfreezes pretrained layers to improve performance across diverse datasets and models.

Contribution

HATL is a novel adaptive transfer learning method that progressively unfreezes pretrained layers based on training behavior, enhancing sign language translation across multiple datasets.

Findings

HATL outperforms traditional transfer learning methods in sign language translation tasks.

ADAT with HATL achieves BLEU-4 improvements of 15.0% on PHOENIX14T and Isharah.

HATL demonstrates robust generalization across three diverse datasets.

Abstract

Sign Language Machine Translation (SLMT) aims to bridge communication between Deaf and hearing individuals. However, its progress is constrained by scarce datasets, limited signer diversity, and large domain gaps between sign motion patterns and pretrained representations. Existing transfer learning approaches in SLMT are static and often lead to overfitting. These challenges call for the development of an adaptive framework that preserves pretrained structure while remaining robust across linguistic and signing variations. To fill this void, we propose a Hierarchical Adaptive Transfer Learning (HATL) framework, where pretrained layers are progressively and dynamically unfrozen based on training performance behavior. HATL combines dynamic unfreezing, layer-wise learning rate decay, and stability mechanisms to preserve generic representations while adapting to sign characteristics. We evaluate HATL on Sign2Text and Sign2Gloss2Text translation tasks using a pretrained ST-GCN++ backbone for feature extraction and the Transformer and an adaptive transformer (ADAT)for translation. To ensure robust multilingual generalization, we evaluate the proposed approach across three datasets: RWTH-PHOENIXWeather-2014 (PHOENIX14T), Isharah, and MedASL. Experimental results show that HATL consistently outperforms traditional transfer learning approaches across tasks and models, with ADAT achieving BLEU-4 improvements of 15.0% on PHOENIX14T and Isharah and 37.6% on MedASL.