Digitizing Nepal's Written Heritage: A Comprehensive HTR Pipeline for Old Nepali Manuscripts

TL;DR

This paper introduces the first comprehensive Handwritten Text Recognition pipeline for Old Nepali, utilizing encoder-decoder models and data techniques to achieve low error rates in recognizing historical manuscripts.

Contribution

It develops an end-to-end HTR pipeline for Old Nepali, including model architectures, decoding strategies, and release of code and configurations for low-resource script recognition.

Findings

Achieved a Character Error Rate of 4.9% on Old Nepali manuscripts.

Systematically explored encoder-decoder architectures and data techniques.

Provided open-source tools to facilitate further research in low-resource HTR.

Abstract

This paper presents the first end-to-end pipeline for Handwritten Text Recognition (HTR) for Old Nepali, a historically significant but low-resource language. We adopt a line-level transcription approach and systematically explore encoder-decoder architectures and data-centric techniques to improve recognition accuracy. Our best model achieves a Character Error Rate (CER) of 4.9\%. In addition, we implement and evaluate decoding strategies and analyze token-level confusions to better understand model behavior and error patterns. Although the evaluation dataset is confidential, we release our training code, model configurations, and evaluation scripts to support further research on HTR for low-resource historical scripts.