Unimore logo AImageLab
Back to the research area

Handwritten Text Recognition on Historical Documents

Handwritten Text Recognition (HTR) aims at automatizing document processing by providing natural language transcriptions of handwritten texts. As such, it plays an important role in automated services, document processing, and Digital Humanities. In this last field, the applications range from the transcription of large document corpora to the analysis of toponyms on ancient maps. Despite Optical Character Recognition (OCR) being a mature and well-established technology, HTR is still a challenging task even when tackled with approaches based on feature learning, especially when it comes to free-layout pages and historical documents.

HTR of Historical documents is a task of great relevance in the context of Digital Humanities since it will allow providing a low-cost semi-automatic transcription of historical manuscripts, thus easing text consultation and automatic keyword search in large collections. Moreover, this task is an interesting and challenging one for modern AI, involving Computer Vision, Natural Language Processing, and Machine Learning.

We are developing advanced Deep Learning models to be applied to the manuscripts in the rich collections in Modena. 

In particular, we are focussing on documents by the Modenese Historian Lodovico Antonio Muratori, due to his cultural relevance for the city of Modena, the vastness of the Muratori corpus preserved in our Digital Libraries (more than 3000 volumes containing letters, minutes, notes; both autographs and not; mainly written in Italian, but also Latin, French, Spanish, German and English), and for the technical challenges that these manuscripts offer (damaged paper, bleeding-trough ink, stains, stroke-out paragraphs). In this respect, we are building a rich line-level HTR dataset of Historical Documents featuring Muratori's manuscripts.

In addition, we are developing strategies to deal with smaller Historical Documents corpora from other authors of Italian and European relevance (data augmentation, data synthesis, domain adaptation).

Check out this presentation of the project for further details: Intelligenza_Artificiale_e_il_riconoscimento_della_scrittura_manoscritta_AIxCH.pdf


1 Cojocaru, Iulian; Cascianelli, Silvia; Baraldi, Lorenzo; Corsini, Massimiliano; Cucchiara, Rita "Watch Your Strokes: Improving Handwritten Text Recognition with Deformable Convolutions" Proceedings of the 25th International Conference on Pattern Recognition, Milan, Italy, pp. 6096 -6103 , 10-15 January 2021, 2021 | DOI: 10.1109/ICPR48806.2021.9412392 Conference
2 Cascianelli, Silvia; Cornia, Marcella; Baraldi, Lorenzo; Piazzi, Maria Ludovica; Schiuma, Rosiana; Cucchiara, Rita "Learning to Read L'Infinito: Handwritten Text Recognition with Synthetic Training Data" Proceedings of the 19th International Conference on Computer Analysis of Images and Patterns, vol. 13053, Virtual, pp. 340 -350 , 27 September - 01 October 2021, 2021 | DOI: 10.1007/978-3-030-89131-2_31 Conference