Ottoman HTR: Recognition of the Ottoman Riqa Font Using Deep Learning Models

This paper presents the preliminary results of a recently trained deep learning model for the recognition of the Riqa font, a common form of Ottoman handwriting, on approximately 250,000 line images. We present the dataset generation process and the training of deep learning models, achieving an overall character error rate of 12.77 % and a word error rate of 40.24 %, setting a strong baseline for Ottoman HTR, on a test set of 85 pages from diverse resources. The results demonstrate the efficacy of our methodology in significantly advancing the state-of-the-art for Ottoman HTR.