























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.
| Comments: | Accepted at ACL 2026 (Main Conference) |
| Subjects: | Machine Learning (cs.LG) |
| Cite as: | arXiv:2512.17111 [cs.LG] |
| (or arXiv:2512.17111v2 [cs.LG] for this version) | |
| https://doi.org/10.48550/arXiv.2512.17111 arXiv-issued DOI via DataCite |
From: Esteban Garces Arias [view email]
[v1]
Thu, 18 Dec 2025 22:43:15 UTC (1,972 KB)
[v2]
Tue, 28 Apr 2026 06:34:12 UTC (1,965 KB)
此内容由惯性聚合(RSS阅读器)自动聚合整理,仅供阅读参考。 原文来自 — 版权归原作者所有。