Adapting TrOCR for Printed Tigrinya Text Recognition: Word-Aware Loss Weighting for Cross-Script Transfer Learning

TL;DR

This paper adapts the TrOCR model for printed Tigrinya script recognition, introducing Word-Aware Loss Weighting to improve accuracy and demonstrating rapid training and public release of resources.

Contribution

It presents the first adaptation of TrOCR for Ge'ez script, with a novel Word-Aware Loss Weighting technique and comprehensive evaluation.

Findings

Achieved 0.22% CER and 97.20% accuracy on synthetic Tigrinya text.

Word-Aware Loss Weighting significantly improves performance, reducing CER by two orders of magnitude.

Full training pipeline completes in under three hours on a consumer GPU.

Abstract

Transformer-based OCR models have shown strong performance on Latin and CJK scripts, but their application to African syllabic writing systems remains limited. We present the first adaptation of TrOCR for printed Tigrinya using the Ge'ez script. Starting from a pre-trained model, we extend the byte-level BPE tokenizer to cover 230 Ge'ez characters and introduce Word-Aware Loss Weighting to resolve systematic word-boundary failures that arise when applying Latin-centric BPE conventions to a new script. The unmodified model produces no usable output on Ge'ez text. After adaptation, the TrOCR-Printed variant achieves 0.22% Character Error Rate and 97.20% exact match accuracy on a held-out test set of 5,000 synthetic images from the GLOCR dataset. An ablation study confirms that Word-Aware Loss Weighting is the critical component, reducing CER by two orders of magnitude compared to vocabulary extension alone. The full pipeline trains in under three hours on a single 8 GB consumer GPU. All code, model weights, and evaluation scripts are publicly released.