Citation: Zhang, C.; Wang, W.;
Zhang, G. Construction of a
Character Dataset for Historical
Uchen Tibetan Documents under
Low-Resource Conditions. Electronics
2022, 11, 3919. https://doi.org/
10.3390/electronics11233919
Academic Editor: Silvia Liberata Ullo
Received: 28 October 2022
Accepted: 25 November 2022
Published: 27 November 2022
Publisher’s Note: MDPI stays neutral
with regard to jurisdictional claims in
published maps and institutional affil-
iations.
Copyright: © 2022 by the authors.
Licensee MDPI, Basel, Switzerland.
This article is an open access article
distributed under the terms and
conditions of the Creative Commons
Attribution (CC BY) license (https://
creativecommons.org/licenses/by/
4.0/).
Article
Construction of a Character Dataset for Historical Uchen
Tibetan Documents under Low-Resource Conditions
Ce Zhang
1,2
, Weilan Wang
1,
* and Guowei Zhang
3
1
Key Laboratory of China’s Ethnic Languages and Information Technology of Ministry of Education,
Northwest Minzu University, Lanzhou 730030, China
2
School of Artificial Intelligence, Chongqing University of Education, Chongqing 400065, China
3
LinkDoc Technology, Beijing 100089, China
* Correspondence: wangweilan@xbmu.edu.cn
Abstract:
The construction of a character dataset is an important part of the research on document
analysis and recognition of historical Tibetan documents. The results of character segmentation
research in the previous stage are presented by coloring the characters with different color values. On
this basis, the characters are annotated, and the character images corresponding to the annotation
are extracted to construct a character dataset. The construction of a character dataset is carried
out as follows: (1) text annotation of segmented characters is performed; (2) the character image is
extracted from the character block based on the real position information; (3) according to the class
of annotated text, the extracted character images are classified to construct a preliminary character
dataset; (4) data augmentation is used to solve the imbalance of classes and samples in the preliminary
dataset; (5) research on character recognition based on the constructed dataset is performed. The
experimental results show that under low-resource conditions, this paper solves the challenges in
the construction of a historical Uchen Tibetan document character dataset and constructs a 610-class
character dataset. This dataset lays the foundation for the character recognition of historical Tibetan
documents and provides a reference for the construction of relevant document datasets.
Keywords:
historical Tibetan documents; character annotation; character extraction; data augmenta-
tion; character recognition
1. Introduction
The content recorded in historical Tibetan documents is extremely rich, involving
all aspects of social development, and has important reference value for the study of
Chinese culture. After much reading and preservation, historical Tibetan documents
have been damaged to varying degrees, and some have even been lost. The protection,
development and utilization of existing historical Tibetan documents have become an
important and urgent topic in ethnic language research. Tibetan is a low-resource language,
which makes it difficult to obtain a large amount of document data, and historical Tibetan
documents are more difficult to obtain, resulting in the late start of relevant research. The
study of historical Tibetan documents began in the 1980s. Since 1991, Kojima et al. have
studied the analysis and recognition research on woodcut Tibetan documents, including
character recognition [
1
,
2
], feature extraction [
3
] and other works. However, these works
did not involve the construction of character datasets. More than a decade later, some
researchers conducted relevant research in layout analysis [
4
], text line segmentation [
5
],
character segmentation [
6
,
7
] on different versions of historical Tibetan document data.
Since 2017, our research group has conducted more analysis and recognition research
on historical Tibetan document images. Han et al. proposed a binarization approach
based on several image processing steps, which achieved high performance in image
binarization [
8
]. Zhao et al. proposed an attention U-Net-based binarization approach
for the historical Tibetan document images [
9
]. Zhou et al., Wang et al. and Hu et al.
Electronics 2022, 11, 3919. https://doi.org/10.3390/electronics11233919 https://www.mdpi.com/journal/electronics
