habib.ibrahim · PyLaia · Published December 24, 2025

Dabbas 1706-1711

Text Recognition

Description

Scope: Recognition of Arabic printed characters from the Aleppo press (1706–1711) Description: The Dabbas-OCR model is a specialized Optical Character Recognition tool trained on the unique typographic output of the Arabic printing press established in Aleppo between 1706 and 1711 by Athanasius Dabbās, then Metropolitan of the Antiochian Church. This model supports the transcription of early Arabic typographic forms modeled on punches carved by Antim the Iberian in Wallachia, specifically for Christian liturgical texts. The typefaces used in Dabbās’s books, though printed, preserve hybrid features reflective of manuscript traditions, including ligatures, overlined diacritics, and stylized letterforms. These types were first created at the princely press of Wallachia under Constantin Brâncoveanu, then transported to Aleppo, where Dabbās installed a press in his metropolitan residence. The printing program included editions of the Four Gospels, Psalter, Horologion, and homiletic works, all produced with Arabic types but firmly rooted in the Byzantine Orthodox liturgical tradition. Model Features: Trained on high-resolution scans of Dabbās’s editions (1706–1711). Recognizes early Arabic print type influenced by manuscript norms. Accommodates multilingual layout (Greek-Arabic) when applicable. Optimized for liturgical and theological vocabulary used in Melkite texts. Designed for integration with TEI/XML encoding tools and textual databases (e.g., e-Cheikho). Historical Note: Athanasius Dabbās's Aleppo press represents the first successful endeavor to print Arabic books with Arabic types in the Middle East. Despite political and religious tensions—including restrictions on Arabic printing in the Ottoman Empire—Dabbās’s initiative succeeded thanks to diplomatic support from Orthodox allies in Wallachia and Ukraine. The press ceased operation in 1711 following the death or fall of his patrons and rising political instability.

Try this model

Dabbas 1706-1711
Use this modelOpen in Transkribus
Very low error rate3.62% CER

Character Error Rate (CER) measures the percentage of characters incorrectly recognised. Lower is better. This model scored 3.62% on its validation set. As a rule of thumb, a CER below 10% is considered good for most handwritten material. This is a larger model trained on diverse material, which generally makes it more robust across different handwriting styles. That said, larger training sets also make it harder to push the CER down further.

Measured on the model's own validation data. Results on your documents may differ depending on handwriting style, document condition, language, and how closely your material resembles the training data.

Words339,447
Lines50,686
Training Pages1,515
Model ID457745
Languages
Arabic
Centuries
18th c.19th c.