EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

Scratch-Trained CRNNs vs. Pretrained Vision Transformers: A Controlled Benchmark of Augmentation and Model Scale in Low-Resource Chagatai HTR

Zhanibek Balabayev, Svitlana Biloshchytska, Beibit Abdikenov, Tomiris Zhaksylyk, Birzhan Ayanbayev, Dimash Rakishev

Information · 01.11.2026 · 17 (11) · 1085

DOI: 10.3390/info17111085 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Information, online erschienen am 01.11.2026. Verfasst von Zhanibek Balabayev und 5 weiteren Autor:innen. Die Arbeit nennt 20 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: Science Committee of the Ministry of Science and Higher Education of the Republic of Kazakhstan.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Digitizing the historical Chagatai manuscripts of the Gunnar Jarring Collection is severely hindered by cursive orthographic complexity and extreme data scarcity. This study establishes a rigorous empirical HTR benchmark, comparing two distinct paradigms under identical, highly constrained conditions: a convolutional recurrent neural network (CRNN) trained from scratch and pretrained Vision-Transformer-based architectures (TrOCR Base and Large). To eliminate intra-page data leakage, a strict page-level 80/10/10 partitioning protocol is enforced across the 2531 manually annotated text lines derived from 230 pages. To guarantee statistical reliability, all configurations are evaluated across three independent random seeds. The fine-tuned TrOCR Large model with architecture-specific data augmentation achieved the highest accuracy, yielding a Character Error Rate (CER) of 9.82 ± 0.05% and a Word Error Rate (WER) of 33.27 ± 0.77%, outperforming the from-scratch CRNN baseline (18.07 ± 0.79% CER) by 8.25% absolute. Our decoupled evaluations reveal that while pretraining provides a robust prior that stabilizes learning, spatial augmentation acts as a crucial regularizer. Our decoupled evaluations reveal that the performance advantage of fine-tuned Vision Transformers over scratch-trained CRNNs stems primarily from the stabilizing prior of pretrained transfer learning, while spatial augmentation acts as a crucial regularizer under extreme data scarcity. Character-level error taxonomies identify systemic failure modes, including deletion biases in Transformers and insertion-heavy “diacritic babbling” in connectionist temporal classification models.

Förderung

  • Science Committee of the Ministry of Science and Higher Education of the Republic of Kazakhstan

Zitieren

APA

Balabayev, Z., Biloshchytska, S., Abdikenov, B., Zhaksylyk, T., Ayanbayev, B., & Rakishev, D. (2026). Scratch-Trained CRNNs vs. Pretrained Vision Transformers: A Controlled Benchmark of Augmentation and Model Scale in Low-Resource Chagatai HTR. Information, 17(11), 1085. https://doi.org/10.3390/info17111085

Vancouver

Balabayev Z, Biloshchytska S, Abdikenov B, Zhaksylyk T, Ayanbayev B, Rakishev D. Scratch-Trained CRNNs vs. Pretrained Vision Transformers: A Controlled Benchmark of Augmentation and Model Scale in Low-Resource Chagatai HTR. Information. 2026;17(11):1085. doi:10.3390/info17111085

BibTeX
@article{Balabayev2026,
  author = {Balabayev, Zhanibek and Biloshchytska, Svitlana and Abdikenov, Beibit and Zhaksylyk, Tomiris and Ayanbayev, Birzhan and Rakishev, Dimash},
  title = {Scratch-Trained CRNNs vs. Pretrained Vision Transformers: A Controlled Benchmark of Augmentation and Model Scale in Low-Resource Chagatai HTR},
  journal = {Information},
  year = {2026},
  volume = {17},
  number = {11},
  pages = {1085},
  doi = {10.3390/info17111085}
}
Kostenlos starten →