An Open-Source Workflow for Handwritten Character Recognition

2022

2020

​A mixed method approach for identification of scribes and authors in handwritten documents will be presented by introducing LiViTo, a tool which combines linguistic insights and computer vision techniques in order to assist researchers in the analysis of handwritten historical documents. This report shows that it is feasible to train neural networks for automatic transcription of handwritten documents and to use these transcriptions as input for further analysis. Hypotheses about scribes can be tested effectively by extracting visual handwriting features and clustering them appropriately. The mixed methods system shows the benefits on both sides linguistics and computer vision. LiViTo was trained with historical Czech texts by 18th century immigrants to Berlin, a total of 564 pages from a corpus of about 5000 handwritten pages without indication of author or scribe. An overview of the development of LiViTo and an introduction into its methodology and its functions will be provided....

Mass-digitization projects in libraries have received world-wide attention. Today a significant amount of historical books, newspapers and journals are already digitized and available online for search and retrieval. The situation is totally different if we look at archives. Though major state and regional archives contain tens or even hundreds of shelf kilometers with most often unique material, the state of digitization lags far behind the library world. One reason may be the absence of a technology that can deal with handwritten documents and convert them into editable text. This paper introduces the concept of a comprehensive Transcription and Recognition Platform (TRP) which may support the development of Handwritten Text Recognition (HTR) technology, based on the assumption that this could best be done within a cooperative model where archives, humanities scholars, computer scientists and the general public are involved and provide specific contributions according to their individual mission and interest.

Hipogrifo. Revista de literatura y cultura del Siglo de Oro, 2023

This is a translation of the article "La Inteligencia Artificial al rescate del Siglo de Oro: transcripción y modernización automática de mil trescientos impresos y manuscritos teatrales". https://www.revistahipogrifo.com/index.php/hipogrifo/article/view/1262 https://doi.org/10.13035/H.2023.11.01.08 Cuéllar, Álvaro. (2023). «La Inteligencia Artificial al rescate del Siglo de Oro. Transcripción y modernización automática de mil trescientos impresos y manuscritos teatrales», Hipogrifo. Revista de literatura y cultura del Siglo de Oro, vol. 11, núm. 1, pp. 101-115, https://doi.org/10.13035/H.2023.11.01.08. A high percentage of theatrical prints and manuscripts from the aurisecular period have never been transcribed in an analogical or, of course, digital format. It is therefore impossible to use these documents to carry out searches of our interest or for the valuable computer analyses (stylometry, topic modelling, sentiment analysis, etc.) that have been developed in recent years. Thanks to Artificial Intelligence (Transkribus) and HTR (Handwritten Text Recognition) techniques, I have trained three models, already public for the research community, capable of transcribing and orthographically modernizing these documents automatically with a high degree of precision: around 97% of success in prints and 91% in manuscripts. Through these models I have been able to process some 1,300 theatrical plays contained in prints and manuscripts from numerous libraries, archives, and other digitized sources. The resulting transcripts are now part of the ETSO project, of the TEXORO search engine and, in addition to being an advanced starting point for careful editing of the texts, they themselves have sufficient quality to be subjected to stylometric analysis, which is yielding authorship attributions of interest.

In Codice Ratio is a research project to study tools and techniques for analyzing the contents of historical documents conserved in the Vatican Secret Archives (VSA). In this paper, we present our eeorts to develop a system to support the transcription of medieval manuscripts. e goal is to provide paleographers with a tool to reduce their eeorts in transcribing large volumes, as those stored in the VSA, producing good transcriptions for signiicant portions of the manuscripts. We propose an original approach based on character segmentation. Our solution is able to deal with the dirty segmentation that inevitably occurs in handwriien documents. We use a convolutional neural network to recognize characters and language models to compose word transcriptions. Our approach requires minimal training eeorts, making the transcription process more scalable as the production of training sets requires a few pages and can be easily crowdsourced. We have conducted experiments on manuscripts from the Vatican Registers, an unreleased corpus containing the correspondence of the popes. With training data produced by 120 high school students, our system has been able to produce good transcriptions that can be used by paleographers as a solid basis to speedup the transcription process at a large scale.

Zenodo (CERN European Organization for Nuclear Research), 2023

Automated text recognition is an ever-present task in the humanities. A multitude of manuscripts and prints from all epochs and cultures have neither been edited nor fully indexed. Even though a considerable amount of them have been digitized, they are usually only accessible as images or PDF files. Machine-actionable transcriptions are often not available, despite being imperative for full-text search, annotation, scholarly editions and text analyses. Consequently, public institutions -including (digital) humanities scholars -require easy to use software solutions which enable them to perform high-quality OCR (Optical Character Recognition) 1 . This proposed full-day workshop introduces the participants to the completely open-source and free of charge software OCR4all ( ) which, unlike other available platforms like Transkribus 2 , eScriptorium 3 , or PERO-OCR 4 project 5 workflows exactly to the specific needs of the material at hand.

Digital Scholarship in the Humanities, 2017

We present a process for cost-effective transcription of cursive handwritten text images that has been tested on a 1 000 pages 17th century book about botanical species. The process comprised two main tasks, namely: (1) preprocessing: page layout analysis, text line detection, and extraction; and (2) transcription of the extracted text line images. Both tasks were carried out with semiautomatic procedures, aimed at incrementally minimizing user correction effort, by means of computer-assisted line detection and interactive handwritten text recognition technologies. The contribution derived from this work is threefold. First, we provide a detailed humansupervised transcription of a relatively large historical handwritten book, ready to be searchable, indexable, and accessible to cultural heritage scholars as well as the general public. Second, we have conducted the first longitudinal study to date on interactive handwriting text recognition, for which we provide a very comprehensive user assessment of the real-world performance of the technologies involved in this work. Third, as a result of this process, we have produced a detailed transcription and document layout information (i.e., high-quality labeled data) ready to be used by researchers working on automated technologies for document analysis and recognition.

2017

Huge amounts of handwritten historical documents are being published by digital libraries world wide. However, for these raw digital images to be really useful, they need to be annotated with informative content. State-of-the-art Handwritten Text Recognition (HTR) approaches require an impressive training effort by expert paleographers. Our contribution is a scalable, end-to-end transcription work-flow – that we call In Codice Ratio – based on fine-grain segmentation of text elements into characters and symbols, with limited training effort. We provide a preliminary evaluation of In Codice Ratio over a corpus of letters by pope Honorii III, stored in the Vatican Secret Archive.

Hipogrifo. Revista de literatura y cultura del Siglo de Oro, 2025

This paper presents a computational approach for detecting the calligraphic footprint of a scribe in a large documentary corpus. The system leverages advances in HTR (Handwritten Text Recognition) techniques, usually employed for automatic transcription, but on this occasion used to locate the specific handwriting of interest when dealing with an extensive collection of texts, in which there may be dozens or even hundreds of different hands. We conducted a control experiment with Lope de Vega (a renowned 17th-century Spanish playwright) and the Transkribus platform (user-friendly for researchers who are not computer specialists), obtaining very accurate results: once trained on Lope’s hand and on two hundred other distinct hands, the system can single out Lope’s handwriting in documents beyond the mod el, with high success rates (accuracy, precision, recall, and F1 scores in the range of 0.95-1.00). These findings pave the way for training models for hands of particular interest (authors, censors, copyists, bureaucrats, actors, etc.) and systematically scanning extant documents in order to detect other instances in which they participated, which could lead to discoveries of historical, literary, and patrimonial significance. Este trabajo presenta una vía computacional para la detección de la huella caligráfica de un escriba en un conjunto documental amplio. El sistema hace uso de los avances en técnicas de HTR (Handwritten Text Recognition), habitual-mente empleadas para la transcripción automática, pero utilizadas en esta ocasión para localizar las manos de nuestro interés al lidiar con un grupo extenso de textos, en el que pueden existir decenas o cientos de manos diferentes. Llevamos a cabo un experimento de control con Lope de Vega (reconocido dramaturgo español del siglo xvii) y la plataforma Transkribus (de fácil uso para los investigadores no es-pecialistas en informática) con resultados muy acertados: el sistema, una vez en-trenado con la mano de Lope y otras doscientas manos diferentes, es capaz de se-ñalar a Lope en documentos fuera del modelo, con unas tasas de acierto elevadas (exactitud, precisión, recall y puntaje F1 en entorno de 0.95-1.00). Estos resultados abren la puerta a entrenar modelos para manos de relevancia (autores, censores, copistas, burócratas, actores, etc.) y a realizar batidas por los documentos conser-vados con el fin de localizar otros testimonios en los que hayan intervenido, lo que puede conducir a hallazgos de valor histórico, literario y patrimonial. Cuéllar, Á., & Boadas, S. (2025). Artificial Intelligence for Calligraphic Writer Identification: The Case of Lope de Vega’s Autographs. Hipogrifo. Revista de literatura y cultura del Siglo de Oro, 13(1), 517–532. https://doi.org/10.13035/H.2025.13.01.36.

Huge amounts of handwritten historical documents are being published by digital libraries world wide. However, for these raw digital images to be really useful, they need to be annotated with informative content. State-of-the-art Handwritten Text Recognition (HTR) approaches require an impressive training effort by expert paleographers. Our contribution is a scalable, end-to-end transcription work-flow – that we call In Codice Ratio – based on fine-grain segmentation of text elements into characters and symbols, with limited training effort. We provide a preliminary evaluation of In Codice Ratio over a corpus of letters by pope Honorii III, stored in the Vatican Secret Archive.