Digitizing Nepal's Written Heritage: A Comprehensive HTR Pipeline for Old Nepali Manuscripts
This work addresses the digitization of Nepal's written heritage, which is an incremental advancement for historical document preservation in a specific domain.
The authors tackled the problem of handwritten text recognition for Old Nepali, a low-resource historical language, by developing an end-to-end pipeline using encoder-decoder architectures and data-centric techniques, achieving a Character Error Rate of 4.9%.
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 behaviour and error patterns. While the dataset we used for evaluation is confidential, we release our training code, model configurations, and evaluation scripts to support further research in HTR for low-resource historical scripts.