Creation and Annotation of a Handwritten Text Database for the Kazakh Language: Methodologies and Preliminary Results

This paper describes the creation and development of a handwritten text database for the Kazakh language, which aims to overcome the lack of publicly available datasets in this sector. While multiple databases exist for handwritten text recognition in other languages, such as IAM for English and RIMES for French, there is no equivalent resource for Kazakh, which employs the Cyrillic alphabet with additional unique characters. This paper explains the systematic process of creating a handwritten Kazakh text library, from gathering data from over 120 writers to annotating text with technologies such as LabelMe. The collection includes 42 Kazakh alphabet letters and more than 75,000 handwritten characters. By introducing this new dataset, we hope to improve research in optical character recognition (OCR) for the Kazakh language and provide the groundwork for future growth in computer vision and text recognition.