The Transkribus Ecosystem

Features

Where Transkribus is used

What Transkribus is used for

Resources

Plans & pricing New

Explore over 100 publicly available AI models for Transkribus. They have been trained and released by the large Transkribus community and the Transkribus team.

Public AI Models

The model was trained with Old Occitan texts, mainly from the 13th and 14th centuries from the Languedoc or Provence. Overall, 7 texts were used for this purpose with a total of about 830 pages and about 190,000 tokens. The following texts were used: 



Las Leys d&#8217;amors (= Bibliothèque municipale de Toulouse, cote 2883/-4), Le roman de Flamenca (=Bibliothèque municipale de Carcassonne, cote 34), and from the Bibliothèque nationale de France (BnF): La vida de sant Enimia (= Arsenal 6355), NAF 11180, NAF 1050, Latin 1139 and Français 846. The model makes independent word segmentations and is capable of resolving the common abbreviations. Punctuation is left as it is in the manuscript (i.e. elided vowels, e.g. in the article, are not replaced by apostrophes). The CER on the training set is 2.6%, on the validation set 3.51%.



<figure class="wp-block-image size-full"><img decoding="async" width="500" height="224" src="https://readcoop.eu/wp-content/uploads/2023/06/Beispiel_Leys_d-amors.png" alt="" class="wp-image-32731" srcset="https://readcoop.eu/wp-content/uploads/2023/06/Beispiel_Leys_d-amors.png 500w, https://readcoop.eu/wp-content/uploads/2023/06/Beispiel_Leys_d-amors-300x134.png 300w, https://readcoop.eu/wp-content/uploads/2023/06/Beispiel_Leys_d-amors-18x8.png 18w" sizes="(max-width: 500px) 100vw, 500px" /></figure>

Old Occitan handwriting

Old Occitan Handwriting

This print model is trained on a selection of six late 19th &#8211; early 20th century Ottoman Turkish periodicals and an Ottoman Turkish dictionary. It was created by the Digital Ottoman Corpora team. We adhered to the “half-transcription” latinization scheme recommended by the Turkish Historical Association for late Ottoman era print material.



<figure class="wp-block-image size-large"><img decoding="async" width="759" height="1024" src="https://readcoop.eu/wp-content/uploads/2023/06/Screen-Shot-2023-06-05-at-6.32.03-PM-759x1024.png" alt="" class="wp-image-32671" srcset="https://readcoop.eu/wp-content/uploads/2023/06/Screen-Shot-2023-06-05-at-6.32.03-PM-759x1024.png 759w, https://readcoop.eu/wp-content/uploads/2023/06/Screen-Shot-2023-06-05-at-6.32.03-PM-222x300.png 222w, https://readcoop.eu/wp-content/uploads/2023/06/Screen-Shot-2023-06-05-at-6.32.03-PM-768x1036.png 768w, https://readcoop.eu/wp-content/uploads/2023/06/Screen-Shot-2023-06-05-at-6.32.03-PM-9x12.png 9w, https://readcoop.eu/wp-content/uploads/2023/06/Screen-Shot-2023-06-05-at-6.32.03-PM-600x810.png 600w, https://readcoop.eu/wp-content/uploads/2023/06/Screen-Shot-2023-06-05-at-6.32.03-PM.png 922w" sizes="(max-width: 759px) 100vw, 759px" /></figure>

Ottoman Turkish Print

OttomanTurkish_Print_1

The model &#8216;Swedish 17th century (Savo, Eastern Finland)&#8217; is trained to read the Gothic handwriting style, also known as the &#8216;German&#8217; handwriting style. It is tailored for 17th-century Swedish handwriting and was conceived as part of Ville-Pekka Kääriäinen&#8217;s doctoral project at the University of Helsinki. The focus of the project was on the 17th century Upper Savonia (Ylä-Savo, Iisalmi/Idensalmi parish, Eastern Finland). Consequently, the model&#8217;s capability to interpret proper nouns (person and place names) might be somewhat limited due to the distinct geographical range of the training data. The model includes training data from various document collections held in the National Archives of both Finland and Sweden:



<ul>
<li>District court (fi: kihlakunnanoikeus, swe: häradsrätten) records related to the Iisalmi parish from 1639–1699 (jurisdictions of Savolax 1639–1650, Kajana friherreskap 1651–1680, Lilla Savolax 1681–1699)</li>



<li>Mentions of Iisalmi parish residents in the lagmansrätt (fi: laamanninoikeus) records from 1643–1699 (Legal district of Karjala, Karelska lagsagan)</li>



<li>Letters sent to Count Per Brahe by local officials, clergy, and townspeople (Skokolstersamlingen, Rydboholmssamlingen)</li>



<li>The complaints of the common people (fi: rahvaanvalitukset, swe: allmogens besvär) from jurisdiction of Lilla-Savolax (fi: Pien-Savo) </li>
</ul>







The image quality of the training data varies. The model was primarily developed before the National Archives of Finland undertook a fresh round of digitization of the court records, necessitating work with low-quality microfilm copies. On the other hand, the material from the Swedish National Archives varies between self-photographed, high-quality digitizations and low-quality digitizations from microfilm. The model adheres closely to the source material in its structure. Monetary units, measurement units, and other abbreviations have been addressed with their inherent logic, albeit without expanding them. For instance, common currency units like the mark and thaler (swe: daler) are depicted by symbols m/m:r or D/D:r, contingent on the context.



The model has been created through substantial personal effort and commitment. It is my hope that it will prove beneficial to others. I am open to collaboration to further develop this model. Please feel free to contact me at: v.kaariainen@gmail.com.



GT: Pages: 1353 (training set) + 147 (validation set) = 1500 pages Words: 472,655 (training set) + 51,613 (validation set) = 524,268 words



<figure class="wp-block-image size-large"><img decoding="async" width="1024" height="790" src="https://readcoop.eu/wp-content/uploads/2023/05/example_swedish17cent-1024x790.jpg" alt="" class="wp-image-32599" srcset="https://readcoop.eu/wp-content/uploads/2023/05/example_swedish17cent-1024x790.jpg 1024w, https://readcoop.eu/wp-content/uploads/2023/05/example_swedish17cent-300x231.jpg 300w, https://readcoop.eu/wp-content/uploads/2023/05/example_swedish17cent-768x592.jpg 768w, https://readcoop.eu/wp-content/uploads/2023/05/example_swedish17cent-1536x1184.jpg 1536w, https://readcoop.eu/wp-content/uploads/2023/05/example_swedish17cent-2048x1579.jpg 2048w, https://readcoop.eu/wp-content/uploads/2023/05/example_swedish17cent-16x12.jpg 16w, https://readcoop.eu/wp-content/uploads/2023/05/example_swedish17cent-600x463.jpg 600w" sizes="(max-width: 1024px) 100vw, 1024px" /></figure>

Swedish 17th century (Savo, Eastern Finland)

This is a first attempt on a model trained for texts in the Romanian Transition Alphabet (1830-1862).



Historical Context and Transcription Principles (by Roxana Patras)



The Romanian transition alphabet is a combination of Cyrillic and Latin characters that was used for printing between approx. 1830, after the Phanariote rulers and the establishment of the Russian protectorate (The Organic Regulation), and approx. 1862, when the first king of The Romanian United Principalities (Moldova and Wallachia) passed a law that constrained typographers to use only Latin glyphs.



Blending in a way that make Romanian printed texts — press, legal documents, and fiction —&nbsp; resemble to the avant-garde Dada poetry, the Cyrillic and Latin letters are not distributed evenly across the 3 decades of transition. Set up through the typographers’ habits and practices and not through convention or linguistic standards, the rules of using Latin instead of Cyrillic are somehow discrete and totally dependent on the location of printing houses. Compared to the Wallachian ones (Bucharest), the Moldavian prints (Iasi) present certain dialectal particularities. Moreover, if we take samples randomly, we can easily notice that the amount of Cyrillic letters is slowly decreasing. However, for some consonants, chiefly palatals, both Cyrillic and Latin letters might be used simultaneously in one and the same text: K/ k&nbsp; — Ч/ч (che); G/g&nbsp; — Г/ г (ghe) —&nbsp; Џ/ џ (Dze); S/s&nbsp; —&nbsp; C/c (es).



In order to train an HTR model for these texts, I have chosen 5 samples that show, before and after 1859, when the 2 Romanian provinces become a country with an official language, the progression from a massive use of Cyrillic letters to an eye-friendly employment, which makes reading more fluent.



The paratextual information characterizes the 5 texts as “original” or “historical” novels. In fact, if we take into account that the first Romanian novel was published in 1845 by a mysterious author signing D.F.B. (Elvira sau amorul fără de sfârşit. Romans original/&nbsp;Elvira or the neverending love. Original Romance), they are regarded by the Romanian literary tradition as a sort of founding pieces:



<div class="wp-block-group is-layout-constrained"><div class="wp-block-group__inner-container">
1.&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;PELIMON Al.,&nbsp;Hoţii şi Hagiul. Roman istoric, Buc., Tip. Sfintei Mitropolii, 1853, 117 p.



2.&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;BOERESCU Costache,&nbsp;Aldo şi Aminta sau Bandiţii, Buc., Tip. Bisericească din Sf. Mitropolie, 1855, 164 p.&nbsp;&nbsp;



3.&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;PELIMON Al.,&nbsp;Jidovul cămătar. Moldova şi Bucovina, Buc., Tip. Stephan Rassidescu, 1863, 292 p.



4.&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;PELIMON Al.,&nbsp;Bucur, istoria fundării Bucureştilor, Buc., Tip. Nationala Iosif Romanov, 1858, 251 p.



5.&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;ARICESCU C.D.,&nbsp;Misterele căsătoriei, I. Bărbatul predestinat, Buc., Tip. Stephan Rassidescu, 1861, 179 p.
</div></div>



As a general rule, Latin capital letters are preferred for writing titles after 1859. The Latin letters Z/ z, M/ m, D/ d, S/ s, T/ t, N/ n, A/ a, I/ i, E/ e, O/ o, Î/ î, U/ u, Ŭ/ ŭ, Ĭ/ ĭ are present from the oldest sampled text (1853), whereas the Cyrillic Х/х (ha), Ш/ ш (sha), Щ/ щ (shcha), Ц/ ц (tze), Џ/ џ (dze), Ч/ ч (che), Ъ/ ъ (ă), П/ п (pe), Р/ р (er), Ж/ ж (zhe), Ф/ф (ef), К/ к (ca), В/ в (ve), Л/ л (el), Г/ г (ghe), Б/ б (be). Among these Cyrillic letters, the first to receive a Latin equivalent are: Ф/ф (ef) → f;&nbsp; Г/ г (ghe) → g; Л/ л (el) → l; Ж/ ж (zhe) → j. At the same time, Р/ р (er), П/ п (pe), Ъ/ ъ (ă), Ч/ ч (che), В/ в (ve), Ш/ ш (sha), Щ/ щ (shcha), Ц/ ц (tse) tend to be maintained until 1862, when some of them they are replaced with glyphs such as “ḑ” (dz), “ş” (sh) and “ț” (tz), which were imported from the Livonian alphabet but have entered the printing circuit only after 1865.



The general guidelines for transcription have been established as follows:



1.&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;Creation of the collection “ALFABET DE TRANZITIE”&nbsp;containing 6 items.



2.&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;Random transcription of initial, middle, and end pages:



2.1.&nbsp;&nbsp;PELIMON Al.,&nbsp;Hoţii şi Hagiul. Roman istoric&nbsp;(1853): pages 6, 7, 8, 52, 53, 54, 76, 77, 78, 79, 80.



2.2.&nbsp;BOERESCU Costache,&nbsp;Aldo şi Aminta sau&nbsp;Bandiţii (1855): pages 8, 9, 10, 36, 37, 38, 114, 115, 116, 148, 150.



2.3.&nbsp;PELIMON Al.,&nbsp;Jidovul cămătar. Moldova şi Bucovina&nbsp;(1863): pages 5, 6, 7, 8, 49, 50, 51, 100, 101, 102.



2.4.&nbsp;PELIMON Al.,&nbsp;Bucur, istoria fundării Bucureştilor&nbsp;(1858): pages 5, 6, 7, 84, 85, 86, 87.



2.5.&nbsp;ARICESCU C.D.,&nbsp;Misterele căsătoriei, I. Bărbatul predestinat&nbsp;(1861): pages 1, 2, 56, 57, 58, 97, 98, 99, 133, 135, 136.



3.&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;Transliteration one-on-one of all Cyrillic letters excepting the situations when K/k stands for the group Ch/ ch (e.g. Бukete&nbsp;→&nbsp;Bukete):



Х/х → H/ h; Ш/ ш → Ș/ ș; Щ/ щ → Șt/ șt; Ц/ ц → Ț/ ț, Ч/ ч → C/ c;



Ъ/ ъ → Ă/ ă; П/ п → P/ p; C/c → S/s;&nbsp; Р/ р → R/ r; Ж/ ж → J/j; Ф/ф → F/ f;



К/ к → C/c; В/ в → V/ v; Л/ л → L/l; Г/ г → G/ g; Б/ б → B/ b; Џ/ џ → G/ g.



4.&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;Customization of the following glyphs:



apostrophe, right double quotation mark, double low-9 quotation mark, Ŭ/ ŭ, Ĭ/ ĭ, á.



<figure class="wp-block-image size-large"><img decoding="async" width="1024" height="348" src="https://readcoop.eu/wp-content/uploads/2023/05/snippet-stranskribus_unboemroman_pantazighica-1024x348.png" alt="" class="wp-image-32566" srcset="https://readcoop.eu/wp-content/uploads/2023/05/snippet-stranskribus_unboemroman_pantazighica-1024x348.png 1024w, https://readcoop.eu/wp-content/uploads/2023/05/snippet-stranskribus_unboemroman_pantazighica-300x102.png 300w, https://readcoop.eu/wp-content/uploads/2023/05/snippet-stranskribus_unboemroman_pantazighica-768x261.png 768w, https://readcoop.eu/wp-content/uploads/2023/05/snippet-stranskribus_unboemroman_pantazighica-18x6.png 18w, https://readcoop.eu/wp-content/uploads/2023/05/snippet-stranskribus_unboemroman_pantazighica-600x204.png 600w, https://readcoop.eu/wp-content/uploads/2023/05/snippet-stranskribus_unboemroman_pantazighica.png 1157w" sizes="(max-width: 1024px) 100vw, 1024px" /></figure>

Romanian Transition Alphabet

RTA2 (Romanian Transition Alphabet)

This first version of a generic model for handwritten Ukrainian (predominantly from the second part of the 19th to the late 20th century) was curated and trained by Aleksej Tikhonov, a researcher of the <a href="http://www.multihtr.uni-freiburg.de">MultiHTR</a> project (Freiburg/Germany, PI: Achim Rabus). Portions of the GT data have kindly been provided by Kateryna Lobuzina of <a href="http://nbuv.gov.ua/node/554">The Vernadskyi National Library of Ukraine</a> (VNLU), containing manuscripts by Taras Shevchenko, as well as Klyment Kvitka’s notices and comments on Ukrainian folklore collected by Lesya Ukrainka. Another part of the GT has kindly been provided by the corpus created by Maria Shvedova, Ruprecht von Waldenfels, Serhij Yaryhin, Andriy Rysin, Vasyl Starko, Tymofij Nikolajenko, et al. (2017-2023): GRAC: General Regionally Annotated Corpus of Ukrainian, Electronic resource: Kyiv, Lviv, Jena (<a href="http://uacorpus.org/">uacorpus.org</a>), which contains a collection of letters prepared by students of Lviv Polytechnic University in 2018-2020 (<a href="http://uacorpus.org/Kyiv/en/gracinfo/rozrobniki">http://uacorpus.org/Kyiv/en/gracinfo/rozrobniki</a>). We thank Mykhailo Kostiv of <a href="https://holodomormuseum.org.ua/en">The National Museum of the Holodomor-Genocide</a> in Kyiv, who provided the training process with a manuscript by Lavrin Nechyporenko from the 1960s. We sincerely thank our partners from Ukraine for the cooperation with MultiHTR on the first generic model of handwritten Ukrainian, despite the ongoing war in their country.



<figure class="wp-block-image size-large"><img decoding="async" width="835" height="1024" src="https://readcoop.eu/wp-content/uploads/2023/05/view_Ukrainka-0005-835x1024.jpg" alt="" class="wp-image-32527" srcset="https://readcoop.eu/wp-content/uploads/2023/05/view_Ukrainka-0005-835x1024.jpg 835w, https://readcoop.eu/wp-content/uploads/2023/05/view_Ukrainka-0005-245x300.jpg 245w, https://readcoop.eu/wp-content/uploads/2023/05/view_Ukrainka-0005-768x941.jpg 768w, https://readcoop.eu/wp-content/uploads/2023/05/view_Ukrainka-0005-1253x1536.jpg 1253w, https://readcoop.eu/wp-content/uploads/2023/05/view_Ukrainka-0005-1671x2048.jpg 1671w, https://readcoop.eu/wp-content/uploads/2023/05/view_Ukrainka-0005-10x12.jpg 10w, https://readcoop.eu/wp-content/uploads/2023/05/view_Ukrainka-0005-600x735.jpg 600w" sizes="(max-width: 835px) 100vw, 835px" /></figure>



Image source: <a href="http://www.nbuv.gov.ua/">Vernadskyi National Library of Ukraine</a>.

Ukrainian generic handwriting

Ukrainian generic handwriting 1

This model is based on 46 different manuscripts from ca. 800 until ca. 1100, mainly written in the Carolingian minuscule. The manuscripts cover different genres such as biblical texts, theological treatises, penitentiary books, histories, mathematical treatises, Canon Law and capitularies and many more. A complete description of the model will be soon published by Tim Geelhaar in AMAD.org. 



<figure class="wp-block-image size-large"><img decoding="async" width="879" height="1024" src="https://readcoop.eu/wp-content/uploads/2023/05/view_EA_ms_clean-0001-879x1024.jpg" alt="" class="wp-image-32520" srcset="https://readcoop.eu/wp-content/uploads/2023/05/view_EA_ms_clean-0001-879x1024.jpg 879w, https://readcoop.eu/wp-content/uploads/2023/05/view_EA_ms_clean-0001-258x300.jpg 258w, https://readcoop.eu/wp-content/uploads/2023/05/view_EA_ms_clean-0001-768x895.jpg 768w, https://readcoop.eu/wp-content/uploads/2023/05/view_EA_ms_clean-0001-1319x1536.jpg 1319w, https://readcoop.eu/wp-content/uploads/2023/05/view_EA_ms_clean-0001-1758x2048.jpg 1758w, https://readcoop.eu/wp-content/uploads/2023/05/view_EA_ms_clean-0001-10x12.jpg 10w, https://readcoop.eu/wp-content/uploads/2023/05/view_EA_ms_clean-0001-600x699.jpg 600w" sizes="(max-width: 879px) 100vw, 879px" /></figure>

Latin &#8211; Carolingian Minuscule

Carolingian Minuscule Model CMM 9th-11th c.

This model is based on the 18th century collection of sermons &#8220;Pridige Od Kerſzhanske Pokore&#8221; by Konrad Branka OFM, written in the Bohorič orthography. It was trained by <a href="marko.kunavar01@gmail.com">Marko Kunavar</a>, Research Centre of the Slovenian Academy of Sciences and Arts. 



<figure class="wp-block-image size-large"><img decoding="async" width="867" height="1024" src="https://readcoop.eu/wp-content/uploads/2023/05/Slovenian-18thcent-867x1024.jpg" alt="" class="wp-image-32448" srcset="https://readcoop.eu/wp-content/uploads/2023/05/Slovenian-18thcent-867x1024.jpg 867w, https://readcoop.eu/wp-content/uploads/2023/05/Slovenian-18thcent-254x300.jpg 254w, https://readcoop.eu/wp-content/uploads/2023/05/Slovenian-18thcent-768x907.jpg 768w, https://readcoop.eu/wp-content/uploads/2023/05/Slovenian-18thcent-1300x1536.jpg 1300w, https://readcoop.eu/wp-content/uploads/2023/05/Slovenian-18thcent-1733x2048.jpg 1733w, https://readcoop.eu/wp-content/uploads/2023/05/Slovenian-18thcent-10x12.jpg 10w, https://readcoop.eu/wp-content/uploads/2023/05/Slovenian-18thcent-600x709.jpg 600w" sizes="(max-width: 867px) 100vw, 867px" /></figure>

Slovenian handwriting 18th century

Slovenian 18th century manuscript

Model for 19th century Danish Gothic handwriting. The material consists of various sources older than 1875. Some material consists of late 18th century and early 19th century handwriting provided by Associate Professor Johan Heinsen. The rest of the training material consists of Parish protocols from Aarhus Stadsarkiv, NæstvedArkiverne, Gentofte Lokalarkiv and Ry Lokalarkiv through the project RETRO. More info can be found on <a href="https://www.retrodigitalisering.dk/">retrodigitalisering.dk</a>.



This model is trained by the archivists Kristian Pindstrup, Mia Gulvad Jørgensen and Anette Larner at Aarhus City Archives.



<figure class="wp-block-image size-large"><img decoding="async" width="736" height="1024" src="https://readcoop.eu/wp-content/uploads/2023/04/Todbjerg-Mejlby-Forhandlingsprotokol-1841-1878_0042-736x1024.jpg" alt="" class="wp-image-32362" srcset="https://readcoop.eu/wp-content/uploads/2023/04/Todbjerg-Mejlby-Forhandlingsprotokol-1841-1878_0042-736x1024.jpg 736w, https://readcoop.eu/wp-content/uploads/2023/04/Todbjerg-Mejlby-Forhandlingsprotokol-1841-1878_0042-216x300.jpg 216w, https://readcoop.eu/wp-content/uploads/2023/04/Todbjerg-Mejlby-Forhandlingsprotokol-1841-1878_0042-768x1068.jpg 768w, https://readcoop.eu/wp-content/uploads/2023/04/Todbjerg-Mejlby-Forhandlingsprotokol-1841-1878_0042-1104x1536.jpg 1104w, https://readcoop.eu/wp-content/uploads/2023/04/Todbjerg-Mejlby-Forhandlingsprotokol-1841-1878_0042-1472x2048.jpg 1472w, https://readcoop.eu/wp-content/uploads/2023/04/Todbjerg-Mejlby-Forhandlingsprotokol-1841-1878_0042-9x12.jpg 9w, https://readcoop.eu/wp-content/uploads/2023/04/Todbjerg-Mejlby-Forhandlingsprotokol-1841-1878_0042-600x835.jpg 600w, https://readcoop.eu/wp-content/uploads/2023/04/Todbjerg-Mejlby-Forhandlingsprotokol-1841-1878_0042-scaled.jpg 1840w" sizes="(max-width: 736px) 100vw, 736px" /></figure>

Danish Gothic handwriting 19th century

19th century Danish Gothic handwriting v.1.1

Trained on a mix of historic printed Hebrew scripts and languages, in the framework of the project DiJeSt. For details: http://dijest.net/gtmodel

Hebrew script languages

DiJeSt 2.0

<figure class="wp-block-image size-large"><img decoding="async" width="1024" height="574" src="https://readcoop.eu/wp-content/uploads/2023/04/early-kurrent-emporer-I-1024x574.png" alt="" class="wp-image-32251" srcset="https://readcoop.eu/wp-content/uploads/2023/04/early-kurrent-emporer-I-1024x574.png 1024w, https://readcoop.eu/wp-content/uploads/2023/04/early-kurrent-emporer-I-300x168.png 300w, https://readcoop.eu/wp-content/uploads/2023/04/early-kurrent-emporer-I-768x430.png 768w, https://readcoop.eu/wp-content/uploads/2023/04/early-kurrent-emporer-I-18x10.png 18w, https://readcoop.eu/wp-content/uploads/2023/04/early-kurrent-emporer-I-600x336.png 600w, https://readcoop.eu/wp-content/uploads/2023/04/early-kurrent-emporer-I.png 1456w" sizes="(max-width: 1024px) 100vw, 1024px" /></figure>



This model can decipher a wide range of Kurrent handwriting in german from the 16th to the early 19th century. Especially in the earlier centuries this model might work better than the usual Transkribus Kurrent model. 



This is a generic model trained on a diverse dataset. Such models provide good results without the need for any extra training work. However, the best results can usually be achieved by training a special model for homogenous material, e. g. texts written by the same person or from a narrow historical period.



Curated by the Transkribus team, this model is occasionally updated with community data for continuous improvement.

Early Kurrent Emperor I.

<figure class="wp-block-image size-large"><img decoding="async" width="1024" height="574" src="https://readcoop.eu/wp-content/uploads/2023/04/german-giant-I-1024x574.png" alt="" class="wp-image-32250" srcset="https://readcoop.eu/wp-content/uploads/2023/04/german-giant-I-1024x574.png 1024w, https://readcoop.eu/wp-content/uploads/2023/04/german-giant-I-300x168.png 300w, https://readcoop.eu/wp-content/uploads/2023/04/german-giant-I-768x430.png 768w, https://readcoop.eu/wp-content/uploads/2023/04/german-giant-I-18x10.png 18w, https://readcoop.eu/wp-content/uploads/2023/04/german-giant-I-600x336.png 600w, https://readcoop.eu/wp-content/uploads/2023/04/german-giant-I.png 1456w" sizes="(max-width: 1024px) 100vw, 1024px" /></figure>



A combination of models containing German language Kurrent and Latin script, also some printed material and smaller amounts of other languages. 



Do not use with a language model, it will not work due to the size of the model. 



This model works well for documents/collections with Kurrent/Sütterlin and Latin script mixed. For only Kurrent or only Latin script the dedicated models might work better, so please test and compare before use. 



Curated by the Transkribus team, this model is occasionally updated with community data for continuous improvement. Based on Transkribus German Handwriting M1.

The German Giant I.

The German Giant I

<figure class="wp-block-image size-large"><img decoding="async" width="1024" height="574" src="https://readcoop.eu/wp-content/uploads/2023/04/dutchess-1024x574.png" alt="" class="wp-image-32219" srcset="https://readcoop.eu/wp-content/uploads/2023/04/dutchess-1024x574.png 1024w, https://readcoop.eu/wp-content/uploads/2023/04/dutchess-300x168.png 300w, https://readcoop.eu/wp-content/uploads/2023/04/dutchess-768x430.png 768w, https://readcoop.eu/wp-content/uploads/2023/04/dutchess-18x10.png 18w, https://readcoop.eu/wp-content/uploads/2023/04/dutchess-600x336.png 600w, https://readcoop.eu/wp-content/uploads/2023/04/dutchess.png 1456w" sizes="(max-width: 1024px) 100vw, 1024px" /></figure>



This public AI model for Transkribus is a powerful combination of four major handwriting recognition models, providing a comprehensive solution for deciphering historical documents from the Low Countries. It covers a wide range of handwritings from the 17th, 18th, and 19th centuries, utilizing a rich dataset contributed by the Nationaal Archief (National Archives of the Netherlands) and the Stadsarchief Amsterdam (Amsterdam City Archives), among other institutions. The model is an amalgamation of the Amsterdam Notarial Super Model, Dutch_XVII_Century, IJsberg, and Dutch Mountains (18th Century), offering an extensive collection of notarial handwritings, chronicles, Dutch East India Company documents, and more.

The Dutchess I.

The Dutchess I

Deep Learning network-based model trained on a dataset of 16th century handwritten Balinese palm leaf manuscript documents.



<figure class="wp-block-image size-large"><img decoding="async" width="1024" height="301" src="https://readcoop.eu/wp-content/uploads/2023/04/Balinese-model-1024x301.jpg" alt="" class="wp-image-32202" srcset="https://readcoop.eu/wp-content/uploads/2023/04/Balinese-model-1024x301.jpg 1024w, https://readcoop.eu/wp-content/uploads/2023/04/Balinese-model-300x88.jpg 300w, https://readcoop.eu/wp-content/uploads/2023/04/Balinese-model-768x226.jpg 768w, https://readcoop.eu/wp-content/uploads/2023/04/Balinese-model-18x5.jpg 18w, https://readcoop.eu/wp-content/uploads/2023/04/Balinese-model-600x176.jpg 600w, https://readcoop.eu/wp-content/uploads/2023/04/Balinese-model.jpg 1225w" sizes="(max-width: 1024px) 100vw, 1024px" /></figure>

Balinese palm-leaf manuscripts 16th century

balinese_v1

English model with some Latin characters, created by the Beyond 2022 project, Trinity College Dublin. Early 17th to late 19th century, 40 hands using Transkribus base model and 750,000 word ground truth. The model is particularly effective on secretary and copperplate texts and can be used as a base model for more cursive scripts.

B2022 English Model

B2022 English Model M4

The HTR model ‘NIOD_WarLet_1935-1950_NoBasemodel’ was trained using 968 &#8216;Ground Truth&#8217; transcriptions of high-resolution scans of various handwritten letters. These letters are all written in Dutch and originate from the period 1935-1950. The training set contains personal correspondence from a wide variety of letter writers (e.g., children, soldiers, Jewish people in hiding). These personal correspondences are all part of the archival collection known as &#8216;247 Correspondentie&#8217; held by the NIOD Institute for War, Holocaust, and Genocide Studies in Amsterdam. This model was created as part of the project &#8216;First-Hand Accounts of War: War letters (1935-1950) from NIOD digitised’. All documents used for training and validation were scanned and transcribed within this project. This project ran from 2020 to 2023 and was funded by the Mondriaan Fund, the Dutch Ministry of Health, Welfare, and Sport, and the NIOD Institute for War, Holocaust, and Genocide Studies in Amsterdam. The &#8216;Ground Truth&#8217; training set is created by project members Annelies van Nispen, Carlijn Keijzer and Milan van Lange. Additional transcription and correction of ‘Ground Truth’ transcriptions was performed under supervision of Muriël Bouman by citizen scientists Hillebrand Verkroost, Bart Cohen, Evelien Bachrach, Marjo Janssens, and Cocky Sietses. The validation set contains a sample of 17 ‘Ground Truth’ transcriptions from various writers and sub-collections. The model is trained using PyLaia HTR, max. 500 epochs (321 epochs trained), learning rate 0.0003. No Base model was used.



<figure class="wp-block-image size-large is-resized"><img decoding="async" src="https://readcoop.eu/wp-content/uploads/2023/02/NL-AsdNIOD_247_1389_0016-817x1024.jpg" alt="" class="wp-image-31811" width="485" height="607" srcset="https://readcoop.eu/wp-content/uploads/2023/02/NL-AsdNIOD_247_1389_0016-817x1024.jpg 817w, https://readcoop.eu/wp-content/uploads/2023/02/NL-AsdNIOD_247_1389_0016-239x300.jpg 239w, https://readcoop.eu/wp-content/uploads/2023/02/NL-AsdNIOD_247_1389_0016-768x963.jpg 768w, https://readcoop.eu/wp-content/uploads/2023/02/NL-AsdNIOD_247_1389_0016-1225x1536.jpg 1225w, https://readcoop.eu/wp-content/uploads/2023/02/NL-AsdNIOD_247_1389_0016-1634x2048.jpg 1634w, https://readcoop.eu/wp-content/uploads/2023/02/NL-AsdNIOD_247_1389_0016-10x12.jpg 10w, https://readcoop.eu/wp-content/uploads/2023/02/NL-AsdNIOD_247_1389_0016-600x752.jpg 600w, https://readcoop.eu/wp-content/uploads/2023/02/NL-AsdNIOD_247_1389_0016-scaled.jpg 2042w" sizes="(max-width: 485px) 100vw, 485px" /></figure>



<figure class="wp-block-image size-large is-resized"><img decoding="async" src="https://readcoop.eu/wp-content/uploads/2023/02/NL-AsdNIOD_247_0038_0165-715x1024.jpg" alt="" class="wp-image-31821" width="486" height="696" srcset="https://readcoop.eu/wp-content/uploads/2023/02/NL-AsdNIOD_247_0038_0165-715x1024.jpg 715w, https://readcoop.eu/wp-content/uploads/2023/02/NL-AsdNIOD_247_0038_0165-209x300.jpg 209w, https://readcoop.eu/wp-content/uploads/2023/02/NL-AsdNIOD_247_0038_0165-768x1100.jpg 768w, https://readcoop.eu/wp-content/uploads/2023/02/NL-AsdNIOD_247_0038_0165-1072x1536.jpg 1072w, https://readcoop.eu/wp-content/uploads/2023/02/NL-AsdNIOD_247_0038_0165-1430x2048.jpg 1430w, https://readcoop.eu/wp-content/uploads/2023/02/NL-AsdNIOD_247_0038_0165-8x12.jpg 8w, https://readcoop.eu/wp-content/uploads/2023/02/NL-AsdNIOD_247_0038_0165-600x859.jpg 600w, https://readcoop.eu/wp-content/uploads/2023/02/NL-AsdNIOD_247_0038_0165-scaled.jpg 1787w" sizes="(max-width: 486px) 100vw, 486px" /></figure>


<div class="wp-block-image">
<figure class="alignleft size-large is-resized"><img decoding="async" src="https://readcoop.eu/wp-content/uploads/2023/02/NL-AsdNIOD_247_0272_0033-653x1024.jpg" alt="" class="wp-image-31820" width="489" height="767" srcset="https://readcoop.eu/wp-content/uploads/2023/02/NL-AsdNIOD_247_0272_0033-653x1024.jpg 653w, https://readcoop.eu/wp-content/uploads/2023/02/NL-AsdNIOD_247_0272_0033-191x300.jpg 191w, https://readcoop.eu/wp-content/uploads/2023/02/NL-AsdNIOD_247_0272_0033-768x1205.jpg 768w, https://readcoop.eu/wp-content/uploads/2023/02/NL-AsdNIOD_247_0272_0033-979x1536.jpg 979w, https://readcoop.eu/wp-content/uploads/2023/02/NL-AsdNIOD_247_0272_0033-1305x2048.jpg 1305w, https://readcoop.eu/wp-content/uploads/2023/02/NL-AsdNIOD_247_0272_0033-8x12.jpg 8w, https://readcoop.eu/wp-content/uploads/2023/02/NL-AsdNIOD_247_0272_0033-600x941.jpg 600w, https://readcoop.eu/wp-content/uploads/2023/02/NL-AsdNIOD_247_0272_0033-scaled.jpg 1631w" sizes="(max-width: 489px) 100vw, 489px" /></figure></div>


<figure class="wp-block-image size-large is-resized"><img decoding="async" src="https://readcoop.eu/wp-content/uploads/2023/02/NL-AsdNIOD_247_0040_0003-710x1024.jpg" alt="" class="wp-image-31818" width="488" height="704" srcset="https://readcoop.eu/wp-content/uploads/2023/02/NL-AsdNIOD_247_0040_0003-710x1024.jpg 710w, https://readcoop.eu/wp-content/uploads/2023/02/NL-AsdNIOD_247_0040_0003-208x300.jpg 208w, https://readcoop.eu/wp-content/uploads/2023/02/NL-AsdNIOD_247_0040_0003-768x1108.jpg 768w, https://readcoop.eu/wp-content/uploads/2023/02/NL-AsdNIOD_247_0040_0003-1064x1536.jpg 1064w, https://readcoop.eu/wp-content/uploads/2023/02/NL-AsdNIOD_247_0040_0003-1419x2048.jpg 1419w, https://readcoop.eu/wp-content/uploads/2023/02/NL-AsdNIOD_247_0040_0003-8x12.jpg 8w, https://readcoop.eu/wp-content/uploads/2023/02/NL-AsdNIOD_247_0040_0003-600x866.jpg 600w, https://readcoop.eu/wp-content/uploads/2023/02/NL-AsdNIOD_247_0040_0003-scaled.jpg 1774w" sizes="(max-width: 488px) 100vw, 488px" /></figure>



<figure class="wp-block-image size-large is-resized"><img decoding="async" src="https://readcoop.eu/wp-content/uploads/2023/02/Beeldmerk_online-liggend-1024x647.jpg" alt="" class="wp-image-31813" width="387" height="244" srcset="https://readcoop.eu/wp-content/uploads/2023/02/Beeldmerk_online-liggend-1024x647.jpg 1024w, https://readcoop.eu/wp-content/uploads/2023/02/Beeldmerk_online-liggend-300x190.jpg 300w, https://readcoop.eu/wp-content/uploads/2023/02/Beeldmerk_online-liggend-768x485.jpg 768w, https://readcoop.eu/wp-content/uploads/2023/02/Beeldmerk_online-liggend-1536x971.jpg 1536w, https://readcoop.eu/wp-content/uploads/2023/02/Beeldmerk_online-liggend-18x12.jpg 18w, https://readcoop.eu/wp-content/uploads/2023/02/Beeldmerk_online-liggend-600x379.jpg 600w, https://readcoop.eu/wp-content/uploads/2023/02/Beeldmerk_online-liggend.jpg 1900w" sizes="(max-width: 387px) 100vw, 387px" /></figure>

NIOD_WarLet_1945-1950_NoBasemodel

NIOD_WarLet_1935-1950_NoBasemodel

This model is based on ground truth from two Old Czech manuscripts: 1) Padeřov Bible (Vienna, Austrian National Library, shelfmark Cod. 1175, 1432–1435), 2) Cistercian Bible (New York, The Morgan Library &amp; Museum, shelfmark MS M.752, 1456), bibles of the third redaction of the Old Czech Bible translation. It can be used on Old Czech manuscripts written in bastard, primarily from the 14th and 15th centuries. The transcription rules were based on semi-diplomatic transcription rules set by Pero OCR and &#8220;Směrnice pro vydávání starších českých textů&#8221; by Jiří Daňhelka. <a href="https://vokabular.ujc.cas.cz/moduly/edicnipoznamka.aspx?id=DanhelkaSmernice">https://vokabular.ujc.cas.cz/moduly/edicnipoznamka.aspx?id=DanhelkaSmernice</a> Abbreviations were tagged and expanded. This model was trained on transcription, where the gaps between words were preserved based on the form in the manuscripts. Produced within the HTR Winter School 2022.



<figure class="wp-block-image size-large"><img decoding="async" width="1024" height="525" src="https://readcoop.eu/wp-content/uploads/2023/02/Old-Czech-Handwriting-model-picture-1024x525.jpg" alt="" class="wp-image-31733" srcset="https://readcoop.eu/wp-content/uploads/2023/02/Old-Czech-Handwriting-model-picture-1024x525.jpg 1024w, https://readcoop.eu/wp-content/uploads/2023/02/Old-Czech-Handwriting-model-picture-300x154.jpg 300w, https://readcoop.eu/wp-content/uploads/2023/02/Old-Czech-Handwriting-model-picture-768x393.jpg 768w, https://readcoop.eu/wp-content/uploads/2023/02/Old-Czech-Handwriting-model-picture-1536x787.jpg 1536w, https://readcoop.eu/wp-content/uploads/2023/02/Old-Czech-Handwriting-model-picture-18x9.jpg 18w, https://readcoop.eu/wp-content/uploads/2023/02/Old-Czech-Handwriting-model-picture-600x307.jpg 600w, https://readcoop.eu/wp-content/uploads/2023/02/Old-Czech-Handwriting-model-picture.jpg 1618w" sizes="(max-width: 1024px) 100vw, 1024px" /></figure>

Old Czech Handwriting (without spaces)

This model is based on ground truth from two Old Czech manuscripts: 1) Padeřov Bible (Vienna, Austrian National Library, shelfmark Cod. 1175, 1432–1435) 2) Cistercian Bible (New York, The Morgan Library &amp; Museum, shelfmark MS M.752, 1456), bibles of the third redaction of the Old Czech Bible translation. It can be used on Old Czech manuscripts written in bastard, primarily from the 14th and 15th centuries. The transcription rules were based on semi-diplomatic transcription rules set by Pero OCR and &#8220;Směrnice pro vydávání starších českých textů&#8221; by Jiří Daňhelka. <a href="https://vokabular.ujc.cas.cz/moduly/edicnipoznamka.aspx?id=DanhelkaSmernice">https://vokabular.ujc.cas.cz/moduly/edicnipoznamka.aspx?id=DanhelkaSmernice</a> Abbreviations were tagged and expanded. This model was trained on transcription, where spaces were created between words based on Jiří Daňhelka&#8217;s semi-diplomatic transcription rules (see above), regardless of the form in the manuscripts. Produced within the HTR Winter School 2022.



<figure class="wp-block-image size-large"><img decoding="async" width="1024" height="525" src="https://readcoop.eu/wp-content/uploads/2023/02/Old-Czech-Handwriting-model-picture-1024x525.jpg" alt="" class="wp-image-31733" srcset="https://readcoop.eu/wp-content/uploads/2023/02/Old-Czech-Handwriting-model-picture-1024x525.jpg 1024w, https://readcoop.eu/wp-content/uploads/2023/02/Old-Czech-Handwriting-model-picture-300x154.jpg 300w, https://readcoop.eu/wp-content/uploads/2023/02/Old-Czech-Handwriting-model-picture-768x393.jpg 768w, https://readcoop.eu/wp-content/uploads/2023/02/Old-Czech-Handwriting-model-picture-1536x787.jpg 1536w, https://readcoop.eu/wp-content/uploads/2023/02/Old-Czech-Handwriting-model-picture-18x9.jpg 18w, https://readcoop.eu/wp-content/uploads/2023/02/Old-Czech-Handwriting-model-picture-600x307.jpg 600w, https://readcoop.eu/wp-content/uploads/2023/02/Old-Czech-Handwriting-model-picture.jpg 1618w" sizes="(max-width: 1024px) 100vw, 1024px" /></figure>

Old Czech Handwriting (with spaces)

Old Czech Handwriting (with spaces), 1st version

Consisting of 18 different notarial handwritings from the 17th century (every handwriting consists of 73-3207 scans training material). In total 11.783 scans training material and 375 scans validation material. Notaries included: Bruijningh, Mathijsz, Schaef, De Winter, Westfrisius, Van der Groe, A. van de Ven, Van Sevenhoven, Des Pommare, Borsselaer, Pondt, De Vos, J. van de Ven, Venkel, Stafmaeker Varlet, Van Breugel, Lock and Doornick. Collection: Amsterdam City Archives.



<figure class="wp-block-image size-large"><img decoding="async" width="650" height="1024" src="https://readcoop.eu/wp-content/uploads/2023/02/view_A27312000218-650x1024.jpg" alt="" class="wp-image-31624" srcset="https://readcoop.eu/wp-content/uploads/2023/02/view_A27312000218-650x1024.jpg 650w, https://readcoop.eu/wp-content/uploads/2023/02/view_A27312000218-191x300.jpg 191w, https://readcoop.eu/wp-content/uploads/2023/02/view_A27312000218-768x1209.jpg 768w, https://readcoop.eu/wp-content/uploads/2023/02/view_A27312000218-976x1536.jpg 976w, https://readcoop.eu/wp-content/uploads/2023/02/view_A27312000218-1301x2048.jpg 1301w, https://readcoop.eu/wp-content/uploads/2023/02/view_A27312000218-8x12.jpg 8w, https://readcoop.eu/wp-content/uploads/2023/02/view_A27312000218-600x945.jpg 600w, https://readcoop.eu/wp-content/uploads/2023/02/view_A27312000218-scaled.jpg 1626w" sizes="(max-width: 650px) 100vw, 650px" /></figure>

Dutch notarial 17th century super model_v2

Notarial 17th Century Super Model_v2

This model was made in a collaboration between Associate Professor Nina Javette Koefoed from Aarhus University and City Archivist Søren Bitsch Christensen and Archivist Kristian Pindstrup of Aarhus City Archives.



It is based on the model 18th Century Danish Prison Records v.5.1. The primary training material consists of 18th Century protocols from the penitentiary on Møn. To increase the universality of the model, further training material has been added from the models &#8220;Various Danish Hands c. 1780 v. 1.13&#8221; by Associate Professor Johan Heinsen from Aalborg University and &#8220;17th Century Norwegian Court Records &#8211; Model Version 5&#8221; by former Senior Advisor Steve Jackson from the National Archival Services of Norway. Over 600.000 words from the 18th century have been used to train the model.

Danish 18th century Administrative Writing

18C Danish Administrative Writing (PyLaia)

General model for 19th century Danish fraktur/gothic printed text. Directories from Aarhus Stadsarkiv with citizen and business information from Aarhus was used as primary source material.



<figure class="wp-block-image size-large"><img decoding="async" width="660" height="1024" src="https://readcoop.eu/wp-content/uploads/2023/01/16-0500_DK-82_1-34-10147_00021-660x1024.jpg" alt="" class="wp-image-31526" srcset="https://readcoop.eu/wp-content/uploads/2023/01/16-0500_DK-82_1-34-10147_00021-660x1024.jpg 660w, https://readcoop.eu/wp-content/uploads/2023/01/16-0500_DK-82_1-34-10147_00021-193x300.jpg 193w, https://readcoop.eu/wp-content/uploads/2023/01/16-0500_DK-82_1-34-10147_00021-768x1191.jpg 768w, https://readcoop.eu/wp-content/uploads/2023/01/16-0500_DK-82_1-34-10147_00021-990x1536.jpg 990w, https://readcoop.eu/wp-content/uploads/2023/01/16-0500_DK-82_1-34-10147_00021-1320x2048.jpg 1320w, https://readcoop.eu/wp-content/uploads/2023/01/16-0500_DK-82_1-34-10147_00021-8x12.jpg 8w, https://readcoop.eu/wp-content/uploads/2023/01/16-0500_DK-82_1-34-10147_00021-600x931.jpg 600w, https://readcoop.eu/wp-content/uploads/2023/01/16-0500_DK-82_1-34-10147_00021-scaled.jpg 1650w" sizes="(max-width: 660px) 100vw, 660px" /></figure>

Danish gothic print 1859-1888 v4

This model was trained as a student project in a master’s program “Digital Humanities” during November 2021 &#8211; January 2022. The text corpus for the model includes books that were published after the Reform of Russian orthography made by Peter I in the following printing houses:  the printing house of the Academy of Sciences in St. Petersburg, the one of the Imperial Moscow University, the one of Vilkovsky and Galchenko, the one of The Land Cadet Corps and some decrees printed in civil script. Training sources are books, scanned by Rusneb (<a href="https://rusneb.ru/" target="_blank" rel="noreferrer noopener">https://rusneb.ru/</a>) and by Google Books.



The model shows good results on Russian language material, but it does not recognize other languages that can occur&nbsp;in texts of this period.



<figure class="wp-block-image size-large"><img decoding="async" width="1024" height="587" src="https://readcoop.eu/wp-content/uploads/2023/01/example1-1024x587.png" alt="" class="wp-image-31472" srcset="https://readcoop.eu/wp-content/uploads/2023/01/example1-1024x587.png 1024w, https://readcoop.eu/wp-content/uploads/2023/01/example1-300x172.png 300w, https://readcoop.eu/wp-content/uploads/2023/01/example1-768x440.png 768w, https://readcoop.eu/wp-content/uploads/2023/01/example1-18x10.png 18w, https://readcoop.eu/wp-content/uploads/2023/01/example1-600x344.png 600w, https://readcoop.eu/wp-content/uploads/2023/01/example1.png 1087w" sizes="(max-width: 1024px) 100vw, 1024px" /></figure>

Russian print XVIII cent.

Russian print XVIII cent PyLaia

PyLaia model created from Ground Truth data resulting from the transcription and manual segmentation of a sample of 193 pages of the Spanish XVIII-XIX press, in particular volumes from &#8220;Diario de Madrid 1788-1825&#8221; (<a href="https://hemerotecadigital.bne.es/hd/card?oid=0001510462" target="_blank" rel="noreferrer noopener">https://hemerotecadigital.bne.es/hd/card?oid=0001510462</a>).



This model has been developed within the CLARA-HD project (<a href="https://clara-nlp.uned.es/home/dh/" target="_blank" rel="noreferrer noopener">https://clara-nlp.uned.es/home/dh</a>) founded by the Spanish Ministery and is valid for &nbsp;automatically transcribing similar Spanish prints of the same period. Manual segmentation is recommended since newspapers usually contain tables and columns. A CER of 1% on validation set has been achieved.



For more information or details please contact Eva Sánchez Salido at <a href="mailto:evasan@lsi.uned.es" target="_blank" rel="noreferrer noopener">evasan@lsi.uned.es</a> or Ana García Serrano at <a href="mailto:agarcia@lsi.uned.es" target="_blank" rel="noreferrer noopener">agarcia@lsi.uned.es</a>. Please cite this model as: Menta, A., Sánchez-Salido, E., &amp; García-Serrano, A. (2022). Transcripción de periódicos históricos: Aproximación CLARA-HD. Proceedings of the Annual Conference of the Spanish Association for Natural Language Processing 2022: Projects and Demonstrations (SEPLN-PD 2022).



<figure class="wp-block-image size-large"><img decoding="async" width="745" height="1024" src="https://readcoop.eu/wp-content/uploads/2023/01/snippet_DM-745x1024.png" alt="" class="wp-image-31435" srcset="https://readcoop.eu/wp-content/uploads/2023/01/snippet_DM-745x1024.png 745w, https://readcoop.eu/wp-content/uploads/2023/01/snippet_DM-218x300.png 218w, https://readcoop.eu/wp-content/uploads/2023/01/snippet_DM-768x1056.png 768w, https://readcoop.eu/wp-content/uploads/2023/01/snippet_DM-9x12.png 9w, https://readcoop.eu/wp-content/uploads/2023/01/snippet_DM-600x825.png 600w, https://readcoop.eu/wp-content/uploads/2023/01/snippet_DM.png 876w" sizes="(max-width: 745px) 100vw, 745px" /></figure>

Spanish print XVIII-XIX

Diario de Madrid 1788-1825

Suomi 1870-1917, version 1 Finnish Handwriting from 1870 to 1917. Created by Ilkka Jokipii (National Archives of Finland), Sami Suodenjoki (University of Tampere) and Maria Niku (Finnish Literature Society). Ground Truth set consist of Finnish Court Records from 1870 to 1917, Citizens letters to the Governor General&#8217;s Office and diaries of Eliel Aspelin-Haapkylä. This is the first version of the model and it will be improved. If you have Finnish transcribed text from models era, contact Ilkka Jokipii (ilkka.jokipii@kansallisarkisto.fi) if you would like to share them to improve the ground truth set.



<figure class="wp-block-image size-large"><img decoding="async" width="1024" height="836" src="https://readcoop.eu/wp-content/uploads/2022/12/KKK1915_Fb1082_19_47_13-1024x836.jpg" alt="" class="wp-image-31212" srcset="https://readcoop.eu/wp-content/uploads/2022/12/KKK1915_Fb1082_19_47_13-1024x836.jpg 1024w, https://readcoop.eu/wp-content/uploads/2022/12/KKK1915_Fb1082_19_47_13-300x245.jpg 300w, https://readcoop.eu/wp-content/uploads/2022/12/KKK1915_Fb1082_19_47_13-768x627.jpg 768w, https://readcoop.eu/wp-content/uploads/2022/12/KKK1915_Fb1082_19_47_13-1536x1254.jpg 1536w, https://readcoop.eu/wp-content/uploads/2022/12/KKK1915_Fb1082_19_47_13-2048x1673.jpg 2048w, https://readcoop.eu/wp-content/uploads/2022/12/KKK1915_Fb1082_19_47_13-15x12.jpg 15w, https://readcoop.eu/wp-content/uploads/2022/12/KKK1915_Fb1082_19_47_13-600x490.jpg 600w" sizes="(max-width: 1024px) 100vw, 1024px" /></figure>

Suomi 1870-1917

This is a model trained on German handwriting of the (early) 19th century Prussian administration. It is based on the Immediatszeitungsberichte of the Rheinprovinz in the years 1816 &amp; 1817. The training material is based on the handwritings of different scribes, texts are mostly in Kurrent.



<figure class="wp-block-image size-large"><img decoding="async" width="676" height="1024" src="https://readcoop.eu/wp-content/uploads/2022/12/Vorschau-Regierungsberichte-676x1024.jpg" alt="" class="wp-image-31115" srcset="https://readcoop.eu/wp-content/uploads/2022/12/Vorschau-Regierungsberichte-676x1024.jpg 676w, https://readcoop.eu/wp-content/uploads/2022/12/Vorschau-Regierungsberichte-198x300.jpg 198w, https://readcoop.eu/wp-content/uploads/2022/12/Vorschau-Regierungsberichte-768x1163.jpg 768w, https://readcoop.eu/wp-content/uploads/2022/12/Vorschau-Regierungsberichte-1014x1536.jpg 1014w, https://readcoop.eu/wp-content/uploads/2022/12/Vorschau-Regierungsberichte-1352x2048.jpg 1352w, https://readcoop.eu/wp-content/uploads/2022/12/Vorschau-Regierungsberichte-8x12.jpg 8w, https://readcoop.eu/wp-content/uploads/2022/12/Vorschau-Regierungsberichte-600x909.jpg 600w, https://readcoop.eu/wp-content/uploads/2022/12/Vorschau-Regierungsberichte.jpg 1608w" sizes="(max-width: 676px) 100vw, 676px" /></figure>

Prussian governmental reports

Preußische Regierungsberichte PyLaia

Model trained by Álvaro Cuéllar for the project ETSO. Estilometría aplicada al Teatro del Siglo de Oro (http://etso.es). It has been trained with theatre prints, mostly 17th and 18th centuries. If you want more information on how to use this model or you want to collaborate somehow, please write to alvarocuellar1995@hotmail.com. If you use this model, please cite -Cuéllar, Álvaro. &#8220;Spanish Golden Age Prints 1.0&#8221;. Transkribus. 2021, and -Cuéllar, Álvaro and Germán Vega García-Luengos. ETSO: Estilometría aplicada al Teatro del Siglo de Oro. 2017-2022. http://etso.es.

Spanish Golden Age Prints 1.0

The Ligorio 0.3 model is based on the handwriting of the Italian antiquarian Pirro Ligorio (Naples 1513 – Ferrara 1583). The training data consists of 57.003 words taken from volume 15, book XVII, letter R, of the alphabetically organised Enciclopedia del Mondo Antico. This Encyclopaedia, written between 1568 and 1583, is now in the State Archives in Turin and can be accessed online on their Digital Library: <a href="https://archiviodistatotorino.beniculturali.it/dbadd/visvol_bibl.php?uid=300146">https://archiviodistatotorino.beniculturali.it/dbadd/visvol_bibl.php?uid=300146</a> [volume 15] The languages used in the text are Italian, Latin and Greek. The transcription criteria used to form the HTR model are: ▪ Normalisation of «V» to «U» except in Latin inscriptions; ▪ Preservation of the diacritical marks and punctuation as used by the author except for the parts in Greek; ▪ Transcription, according to the grammatical rules of the Italian language, when the use of capital and small caps are not distinguishable. Due to an error that occurs when visualizing the Roman sestertius and denarius in Latin inscriptions, used by Pirro Ligorio, the symbols were replaced with others that the author would have never used in his text. The model was trained accordingly and the symbols were chosen from the Astronomical group already present on the virtual keyboard in Transkribus: Roman sestertius sign (U+10198) ➛ ☿ (U+263F Mercury) Roman denarius sign (U+10196) ➛ ♀(U+2640 Female sign). This model was created as part of a Phd project and is part of a wider project named Ligorio Digitale. For more information regarding the background of the model please visit: <a href="https://limes.cfs.unipi.it/lidiws/">https://limes.cfs.unipi.it/lidiws/</a> [Italian]. Citation information (CMSO, website content): Agostini, Giorgia. “The Ligorio 0.3 model, Free Public AI Model for Handwritten Text Recognition with Transkribus”. Last modified [date]. [model link]

Ligorio 0.3

Ligorio 0.3 PyL

HTR model for the handwriting (German Kurrent) of the journalist Paul Goldmann (1865–1935). The model was trained with letters from 1889 to 1896 by Martin Anton Müller and Laura Untner (Austrian Academy of Sciences).







<figure class="wp-block-image size-large"><img decoding="async" width="631" height="1024" src="https://readcoop.eu/wp-content/uploads/2022/10/Paul-Goldmann_German-Kurrent_1889–96_V1-631x1024.jpg" alt="" class="wp-image-30730" srcset="https://readcoop.eu/wp-content/uploads/2022/10/Paul-Goldmann_German-Kurrent_1889–96_V1-631x1024.jpg 631w, https://readcoop.eu/wp-content/uploads/2022/10/Paul-Goldmann_German-Kurrent_1889–96_V1-185x300.jpg 185w, https://readcoop.eu/wp-content/uploads/2022/10/Paul-Goldmann_German-Kurrent_1889–96_V1-768x1245.jpg 768w, https://readcoop.eu/wp-content/uploads/2022/10/Paul-Goldmann_German-Kurrent_1889–96_V1-947x1536.jpg 947w, https://readcoop.eu/wp-content/uploads/2022/10/Paul-Goldmann_German-Kurrent_1889–96_V1-1263x2048.jpg 1263w, https://readcoop.eu/wp-content/uploads/2022/10/Paul-Goldmann_German-Kurrent_1889–96_V1-7x12.jpg 7w, https://readcoop.eu/wp-content/uploads/2022/10/Paul-Goldmann_German-Kurrent_1889–96_V1-600x973.jpg 600w, https://readcoop.eu/wp-content/uploads/2022/10/Paul-Goldmann_German-Kurrent_1889–96_V1-scaled.jpg 1579w" sizes="(max-width: 631px) 100vw, 631px" /></figure>

Paul-Goldmann_German-Kurrent_1889–96_V1

This is a combined model of two Portuguese sources housed at the Portuguese National Archive Torre do Tombo and State Archive of Bahia, Brazil.



There are handwritten and printed scripts of the Inquisition from Torre do Tombo together with the Notarial Books of Salvador da Bahia. To make the model even more general, a printed script from the middle of the 17th Century was added. Some documents are damaged.



This is the first attempt to create a general Portuguese Model with 64,842 words. Two different projects are working on the collections independently. As they progress in their work, new general models will be created, to attend to other collections.



<figure class="wp-block-image size-large"><img decoding="async" width="1024" height="770" src="https://readcoop.eu/wp-content/uploads/2022/10/Screenshot-2022-10-13-at-12.26.32-1024x770.png" alt="" class="wp-image-30688" srcset="https://readcoop.eu/wp-content/uploads/2022/10/Screenshot-2022-10-13-at-12.26.32-1024x770.png 1024w, https://readcoop.eu/wp-content/uploads/2022/10/Screenshot-2022-10-13-at-12.26.32-300x226.png 300w, https://readcoop.eu/wp-content/uploads/2022/10/Screenshot-2022-10-13-at-12.26.32-768x577.png 768w, https://readcoop.eu/wp-content/uploads/2022/10/Screenshot-2022-10-13-at-12.26.32-16x12.png 16w, https://readcoop.eu/wp-content/uploads/2022/10/Screenshot-2022-10-13-at-12.26.32-600x451.png 600w, https://readcoop.eu/wp-content/uploads/2022/10/Screenshot-2022-10-13-at-12.26.32.png 1362w" sizes="(max-width: 1024px) 100vw, 1024px" /></figure>

General Portuguese

General Portuguese M1

This is a combined model of ground truth of different charter and book scripts from a variety of projects and institutions, aiming at building a generic model for Latin scripts of the Middle Ages.&nbsp; It is mainly based on documents from the project CREMMA Manuscrits médiévaux latins, HIMANIS (CNRS), Itinera Nova (Stadsarchief Leuven), and Charters and Records of Königsfelden (Universität Zürich).



CREMMA Manuscrits médiévaux latins has been produced by Clérice, Thibault and Chagué, Alix and Vlachou Efstathiou, Malamatenia. It is licensed under a CC-BY 4.0 license. URL: <a href="https://github.com/HTR-United/CREMMA-Medieval-LAT" target="_blank" rel="noreferrer noopener">https://github.com/HTR-United/CREMMA-Medieval-LAT</a>



HIMANIS is partially published as HIMANIS Guérin produced by Stutzmann, Dominique; Hamel, Sébastien; Kernier, Iseut de; Mühlberger, Günter; Hackl, Günter. Licensed under a CC-BY 4.0 license. DOI: 10.5281/zenodo.5535306



Charters and Records of Königsfelden Abbey and Bailiwick (1308-1662) has been produced by Halter-Pernet, Colette; Teuscher, Simon; Hodel, Tobias; Barwitzki, Lukas; Egloff, Salome; Henggeler, Fabian; Nadig, Michael; Steinmann, Anina; Stettler, Sabine; Prada Ziegler, Ismail. Licensed under a CC-BY 4.0 license. DOI: 10.5281/zenodo.5179361 



<figure class="wp-block-image size-full"><img decoding="async" width="732" height="515" src="https://readcoop.eu/wp-content/uploads/2021/03/Charter-Scripts.jpg" alt="" class="wp-image-21203" srcset="https://readcoop.eu/wp-content/uploads/2021/03/Charter-Scripts.jpg 732w, https://readcoop.eu/wp-content/uploads/2021/03/Charter-Scripts-300x211.jpg 300w, https://readcoop.eu/wp-content/uploads/2021/03/Charter-Scripts-16x12.jpg 16w, https://readcoop.eu/wp-content/uploads/2021/03/Charter-Scripts-600x422.jpg 600w, https://readcoop.eu/wp-content/uploads/2021/03/Charter-Scripts-85x60.jpg 85w, https://readcoop.eu/wp-content/uploads/2021/03/Charter-Scripts-128x90.jpg 128w" sizes="(max-width: 732px) 100vw, 732px" /></figure>

Medieval_Scripts_M2.4

The model is based on a selection of late medieval German manuscripts from the fifteenth and sixteenth centuries. With a geographical focus on Austria and the Hanseatic Baltic region, both Middle Low and Early New High German texts are represented in the training material. Among the fonts used, Gothic cursive, Bastarda, and early Kurrent are most common.



The training data set is composed of approximately 77.000 words. The ground truth was compiled from several projects currently in progress at various institutions. The aim is to significantly expand and diversify this in the future to train a generic model. The current model therefore has a prototype status, with a CER of 5.60% on a validation set. 



Contact: j.helmchen@fu-berlin.de 



Image: WStLA 2.1.2.1.B1.6b.2 &#8211; Gewährbuch C, fol. 292r (1420)



<figure class="wp-block-image size-large"><img decoding="async" width="1024" height="568" src="https://readcoop.eu/wp-content/uploads/2022/09/WStLA_Grundbuch_Stadt_Wien_B1_6b_081-1024x568.jpg" alt="" class="wp-image-30344" srcset="https://readcoop.eu/wp-content/uploads/2022/09/WStLA_Grundbuch_Stadt_Wien_B1_6b_081-1024x568.jpg 1024w, https://readcoop.eu/wp-content/uploads/2022/09/WStLA_Grundbuch_Stadt_Wien_B1_6b_081-300x166.jpg 300w, https://readcoop.eu/wp-content/uploads/2022/09/WStLA_Grundbuch_Stadt_Wien_B1_6b_081-768x426.jpg 768w, https://readcoop.eu/wp-content/uploads/2022/09/WStLA_Grundbuch_Stadt_Wien_B1_6b_081-1536x852.jpg 1536w, https://readcoop.eu/wp-content/uploads/2022/09/WStLA_Grundbuch_Stadt_Wien_B1_6b_081-2048x1136.jpg 2048w, https://readcoop.eu/wp-content/uploads/2022/09/WStLA_Grundbuch_Stadt_Wien_B1_6b_081-18x10.jpg 18w, https://readcoop.eu/wp-content/uploads/2022/09/WStLA_Grundbuch_Stadt_Wien_B1_6b_081-600x333.jpg 600w" sizes="(max-width: 1024px) 100vw, 1024px" /></figure>

Generic Model 15th-16th century German (prototype)

This HTR+ model was trained for various Devanagari print types based on approximately 26,000 words. The training material is comprised of letterpress printings from the Naval Kishore Press (Lakhnau, North India) from the late 19th and early 20th century in the Hindi, Sanskrit, Braj Bhasha and Awadhi languages. The model was created as part of the project &#8220;Naval Kishore Press &#8211; digital&#8221;, which was financially supported by the German Research Foundation (DFG) until 2021. Dieses HTR+ Modell wurde für verschiedene Devanagari-Drucktypen auf der Grundlage von ca. 26.000 Wörtern trainiert. Beim Trainingsmaterial handelt es sich um Buchdrucke der Naval Kishore Press (Lakhnau, Nordindien) aus dem späten 19. und frühen 20. Jahrhundert in den Sprachen Hindi, Sanskrit, Braj Bhasha und Awadhi. Das Modell wurde im Rahmen des Projekts &#8220;Naval Kishore Press &#8211; digital&#8221; erstellt, das von der Deutschen Forschungsgemeinschaft (DFG) bis 2021 finanziell unterstützt wurde. <a href="https://digi.ub.uni-heidelberg.de/sammlungen/suedasien/navalkishore.html" target="_blank" rel="noreferrer noopener">https://digi.ub.uni-heidelberg.de/sammlungen/suedasien/navalkishore.html</a> 







<figure class="wp-block-image size-full"><img decoding="async" width="637" height="905" src="https://readcoop.eu/wp-content/uploads/2022/09/jnanadasa1895.png" alt="" class="wp-image-30040" srcset="https://readcoop.eu/wp-content/uploads/2022/09/jnanadasa1895.png 637w, https://readcoop.eu/wp-content/uploads/2022/09/jnanadasa1895-211x300.png 211w, https://readcoop.eu/wp-content/uploads/2022/09/jnanadasa1895-8x12.png 8w, https://readcoop.eu/wp-content/uploads/2022/09/jnanadasa1895-600x852.png 600w" sizes="(max-width: 637px) 100vw, 637px" /></figure>

Devanagari mixed M1A

The model «Russian print of the 18 c. (V. Okorokov&#8217;s Printing House)» was created during the course of the project «Plants and people in the Russian Empire of the 18th century: knowledge and practice distribution among social classes» at the European University at St. Petersburg. At the first stage, several scientific treatises published at the V. Okorokov’s Printing House (at the Moscow State University) were chosen as a starting point for transcribing. The books were printed in Russian, but some scientific notions and terms were supplied with Latin equivalents.



The model also shows good results on the other printed Russian texts of the 18th c., and can be used as a base model to speed up the creation of a general model for Russian print of the 18th c., including every specific printing house.



Training sources are books, scanned by Google Books (include more than 82,000 words):&nbsp;&nbsp;



<a href="https://www.google.ru/books/edition/Георгия_Готлиба_Рихте/_7ZiAAAAcAAJ?hl=ru&amp;gbpv=0">https://www.google.ru/books/edition/Георгия_Готлиба_Рихте/_7ZiAAAAcAAJ?hl=ru&amp;gbpv=0</a>



<a href="https://www.google.ru/books/edition/О_дѣйствиях_и_силѣ_шал/v65iAAAAcAAJ?hl=ru&amp;gbpv=0">https://www.google.ru/books/edition/О_дѣйствиях_и_силѣ_шал/v65iAAAAcAAJ?hl=ru&amp;gbpv=0</a> 



<figure class="wp-block-image size-full"><img decoding="async" width="570" height="327" src="https://readcoop.eu/wp-content/uploads/2022/09/Russian-print-of-the-18-c.-V.-Okorokovs-Printing-House.png" alt="" class="wp-image-30015" srcset="https://readcoop.eu/wp-content/uploads/2022/09/Russian-print-of-the-18-c.-V.-Okorokovs-Printing-House.png 570w, https://readcoop.eu/wp-content/uploads/2022/09/Russian-print-of-the-18-c.-V.-Okorokovs-Printing-House-300x172.png 300w, https://readcoop.eu/wp-content/uploads/2022/09/Russian-print-of-the-18-c.-V.-Okorokovs-Printing-House-18x10.png 18w" sizes="(max-width: 570px) 100vw, 570px" /></figure>

Russian print of the 18 c. (V. Okorokov&#8217;s Printing House)

Russian print of the 18 c. (V. Okorokov's Printing House)

This model has a train set of 203189 words, stemming from the Zurich Council Minutes (Ratsmanuale) during the 18th century and encompassing numerous scribes from the city chancery in this period. The Ratsmanuale contain the minutes of the meetings of the authorities from the old city-state of Zurich until 1798.



<figure class="wp-block-image size-large"><img decoding="async" width="338" height="1024" src="https://readcoop.eu/wp-content/uploads/2022/08/Staatsarchiv-Zurich-338x1024.jpg" alt="" class="wp-image-29840" srcset="https://readcoop.eu/wp-content/uploads/2022/08/Staatsarchiv-Zurich-338x1024.jpg 338w, https://readcoop.eu/wp-content/uploads/2022/08/Staatsarchiv-Zurich-99x300.jpg 99w, https://readcoop.eu/wp-content/uploads/2022/08/Staatsarchiv-Zurich-768x2325.jpg 768w, https://readcoop.eu/wp-content/uploads/2022/08/Staatsarchiv-Zurich-507x1536.jpg 507w, https://readcoop.eu/wp-content/uploads/2022/08/Staatsarchiv-Zurich-4x12.jpg 4w, https://readcoop.eu/wp-content/uploads/2022/08/Staatsarchiv-Zurich-600x1816.jpg 600w, https://readcoop.eu/wp-content/uploads/2022/08/Staatsarchiv-Zurich-scaled.jpg 846w" sizes="(max-width: 338px) 100vw, 338px" /></figure>

Swiss German Kurrent 18th Century

This Dutch mixed roman/gothic model has been trained from a ground truth set of 100 images of 17th century Dutch newspaper material from the &#8220;Couranten Corpus&#8221;, cf. <a href="https://couranten.ivdnt.org/corpus-frontend/couranten/about">https://couranten.ivdnt.org/corpus-frontend/couranten/about</a>.



The corpus consists of ground truth transcriptions of newspapers digitized by the Dutch National Library.



The model was trained with default parameters, with a 10% validation set.



<figure class="wp-block-image size-full"><img decoding="async" width="950" height="631" src="https://readcoop.eu/wp-content/uploads/2022/08/dutch-newspapers.png" alt="" class="wp-image-29760" srcset="https://readcoop.eu/wp-content/uploads/2022/08/dutch-newspapers.png 950w, https://readcoop.eu/wp-content/uploads/2022/08/dutch-newspapers-300x199.png 300w, https://readcoop.eu/wp-content/uploads/2022/08/dutch-newspapers-768x510.png 768w, https://readcoop.eu/wp-content/uploads/2022/08/dutch-newspapers-18x12.png 18w, https://readcoop.eu/wp-content/uploads/2022/08/dutch-newspapers-600x399.png 600w" sizes="(max-width: 950px) 100vw, 950px" /></figure>

Dutch newspapers 17th century

Italian handwriting, 16th-19th century.



This is a generic model trained on a diverse dataset. Such models provide good results without the need for any extra training work. However, the best results can usually be achieved by training a special model for homogenous material, e. g. texts written by the same person or from a narrow historical period.

Italian &#8211; General Model

The model is based on Danish-Norwegian transcripts from the 20th century of “Samlinger til kildeutgivelse, Haugianerbrev” (1760–1842) at the National Archives of Norway. The documents have running text with the occasional handwritten notes and corrections.







<figure class="wp-block-image size-large"><img decoding="async" width="730" height="1024" src="https://readcoop.eu/wp-content/uploads/2022/07/haugianerbrev_transcripts-730x1024.jpg" alt="" class="wp-image-29479" srcset="https://readcoop.eu/wp-content/uploads/2022/07/haugianerbrev_transcripts-730x1024.jpg 730w, https://readcoop.eu/wp-content/uploads/2022/07/haugianerbrev_transcripts-214x300.jpg 214w, https://readcoop.eu/wp-content/uploads/2022/07/haugianerbrev_transcripts-768x1077.jpg 768w, https://readcoop.eu/wp-content/uploads/2022/07/haugianerbrev_transcripts-1095x1536.jpg 1095w, https://readcoop.eu/wp-content/uploads/2022/07/haugianerbrev_transcripts-9x12.jpg 9w, https://readcoop.eu/wp-content/uploads/2022/07/haugianerbrev_transcripts-600x842.jpg 600w, https://readcoop.eu/wp-content/uploads/2022/07/haugianerbrev_transcripts.jpg 1460w" sizes="(max-width: 730px) 100vw, 730px" /></figure>

Danish-Norwegian Typewritten Transcripts 1760–1842

Haugianerbrev Transcripts

The model is based on Norwegian typewritten interrogation reports from Kjesäter, which was a reception center for Norwegian refugees in Sweden during the Second World War. 



The data is sensitive, so only the model’s neural network is made publicly available.



The data has archival reference “S-1725 &#8211; Utenriksstasjonene, Legasjonen/Ambassaden i Stockholm” (1940–1945) at the National Archives of Norway.

Norwegian Typewriting 1940–1945

Kjesäter Interrogation Reports

The model is based on two Danish-Norwegian partition protocols for Sunnhordland, Norway. Both protocols have handwritten running text. The archival references are “SAB, Sunnhordland Sorenskriverembete, F/Fb/Fbb/L008, Storskifteprotokoll Etne M.fl.” (1874-1882) and “SAB, Sunnhordland Sorenskriverembete, F/Fb/Fbb/L009, Storskifteprotokoll Etne M.fl.” (1883-1891) at the National Archives of Norway. 



<figure class="wp-block-image size-large"><img decoding="async" width="647" height="1024" src="https://readcoop.eu/wp-content/uploads/2022/07/sunnhordland_partition_protocols-647x1024.jpg" alt="" class="wp-image-29469" srcset="https://readcoop.eu/wp-content/uploads/2022/07/sunnhordland_partition_protocols-647x1024.jpg 647w, https://readcoop.eu/wp-content/uploads/2022/07/sunnhordland_partition_protocols-189x300.jpg 189w, https://readcoop.eu/wp-content/uploads/2022/07/sunnhordland_partition_protocols-768x1216.jpg 768w, https://readcoop.eu/wp-content/uploads/2022/07/sunnhordland_partition_protocols-970x1536.jpg 970w, https://readcoop.eu/wp-content/uploads/2022/07/sunnhordland_partition_protocols-8x12.jpg 8w, https://readcoop.eu/wp-content/uploads/2022/07/sunnhordland_partition_protocols-600x950.jpg 600w, https://readcoop.eu/wp-content/uploads/2022/07/sunnhordland_partition_protocols.jpg 1293w" sizes="(max-width: 647px) 100vw, 647px" /></figure>

Danish-Norwegian Handwriting 1874–1891

Sunnhordland Partition Protocols

Prof. Vladimir Polomac (University of Kragujevac, Serbia) created the Dionisio 2.0. generic model for automatic recognition of Serbian Church Slavonic printed books of the 15th-17th centuries. The model was created based on the materials from various Serbian printing houses of the 15th-17th centuries (Cetinje, Venice, Goražde, Mileševa, Belgrade and Mrkša&#8217;a Church). The percentage of incorrectly recognized characters on the validation set is 2.4%. The process of creating the model within CITLab HTR+ engine, its quantitative and qualitative performance are described in detail in a separate <a href="https://nl.ijs.si/jtdh22/pdf/JTDH2022_Polomac_Serbian-Early-Printed-Books-Towards-Generic-Model-for-Automatic-Text.pdf">paper</a>. While the models based on the CITlabHTR+ technology are no longer available in Transkribus, the Dionisio 2.0. model has been retrained with PyLaia technology with the almost identical performances.



<figure class="wp-block-image size-large"><img decoding="async" width="702" height="1024" src="https://readcoop.eu/wp-content/uploads/2022/05/Prayer-Book-1538-1540-folio-12a-702x1024.jpg" alt="" class="wp-image-29073" srcset="https://readcoop.eu/wp-content/uploads/2022/05/Prayer-Book-1538-1540-folio-12a-702x1024.jpg 702w, https://readcoop.eu/wp-content/uploads/2022/05/Prayer-Book-1538-1540-folio-12a-206x300.jpg 206w, https://readcoop.eu/wp-content/uploads/2022/05/Prayer-Book-1538-1540-folio-12a-768x1120.jpg 768w, https://readcoop.eu/wp-content/uploads/2022/05/Prayer-Book-1538-1540-folio-12a-8x12.jpg 8w, https://readcoop.eu/wp-content/uploads/2022/05/Prayer-Book-1538-1540-folio-12a-600x875.jpg 600w, https://readcoop.eu/wp-content/uploads/2022/05/Prayer-Book-1538-1540-folio-12a.jpg 913w" sizes="(max-width: 702px) 100vw, 702px" /></figure>

Dionisio 2.0

Dionisio 2.0.

This is a general model for Hungarian handwriting from the late 19th to the first quarter of the 20th.



The Digital Humanities Centre of the National Széchényi Library created the model. It is based on the corpus of the correspondence between József Kiss (1843–1921), a Hungarian poet and editor and his various connections.



The training data contains his professional and personal letters written throughout his life. The manuscripts used in the current model are located in the Petőfi Literary Museum, Budapest.



More manuscripts of the same correspondence are available and under processing at the National Széchényi Library. We expect to further enlarge the model later on when their process is complete. 



<figure class="wp-block-image size-large"><img decoding="async" width="809" height="1024" src="https://readcoop.eu/wp-content/uploads/2022/05/PKELV.1344_a-809x1024.jpg" alt="" class="wp-image-28629" srcset="https://readcoop.eu/wp-content/uploads/2022/05/PKELV.1344_a-809x1024.jpg 809w, https://readcoop.eu/wp-content/uploads/2022/05/PKELV.1344_a-237x300.jpg 237w, https://readcoop.eu/wp-content/uploads/2022/05/PKELV.1344_a-768x972.jpg 768w, https://readcoop.eu/wp-content/uploads/2022/05/PKELV.1344_a-1214x1536.jpg 1214w, https://readcoop.eu/wp-content/uploads/2022/05/PKELV.1344_a-1619x2048.jpg 1619w, https://readcoop.eu/wp-content/uploads/2022/05/PKELV.1344_a-9x12.jpg 9w, https://readcoop.eu/wp-content/uploads/2022/05/PKELV.1344_a-600x759.jpg 600w, https://readcoop.eu/wp-content/uploads/2022/05/PKELV.1344_a-scaled.jpg 2023w" sizes="(max-width: 809px) 100vw, 809px" /></figure>

Hungarian handwriting 19th–20th cent.

Hungarian handwriting 19th–20th Century

Yiddish Handwriting model, prepared by the project team of <a href="https://www.dybbuk.co/">https://www.dybbuk.co/</a>. Ground truth data consists mainly of play manuscripts from late 19th and early 20th century written in New York and Europe. Also included are several Yiddish letters corrected by participants of Yiddish Transkribus workshop <a href="https://ingeveb.org/pedagogy/the-dybbuk-model-a-new-yiddish-handwriting-text-recognition-tool">https://ingeveb.org/pedagogy/the-dybbuk-model-a-new-yiddish-handwriting-text-recognition-tool</a> .



<figure class="wp-block-image size-large"><img decoding="async" width="1024" height="709" src="https://readcoop.eu/wp-content/uploads/2022/04/dybbuk3photo-1024x709.jpg" alt="" class="wp-image-28364" srcset="https://readcoop.eu/wp-content/uploads/2022/04/dybbuk3photo-1024x709.jpg 1024w, https://readcoop.eu/wp-content/uploads/2022/04/dybbuk3photo-300x208.jpg 300w, https://readcoop.eu/wp-content/uploads/2022/04/dybbuk3photo-768x532.jpg 768w, https://readcoop.eu/wp-content/uploads/2022/04/dybbuk3photo-1536x1064.jpg 1536w, https://readcoop.eu/wp-content/uploads/2022/04/dybbuk3photo-2048x1419.jpg 2048w, https://readcoop.eu/wp-content/uploads/2022/04/dybbuk3photo-18x12.jpg 18w, https://readcoop.eu/wp-content/uploads/2022/04/dybbuk3photo-600x416.jpg 600w" sizes="(max-width: 1024px) 100vw, 1024px" /></figure>

The Dybbuk for Yiddish Handwriting

This model was developed by the L&#8217;Dor V&#8217;Dor Foundation&#8217;s (LDVDF) Documentation of Jewish Records Worldwide (DoJR) project in partnership with the International Association of Jewish Genealogical Societies (IAJGS). It is based on scans of documents from Congress Poland, Ukraine and Russia 1914-1968 and works particularly well with civil records from Congress Poland. 



<figure class="wp-block-image size-full is-resized"><img decoding="async" src="https://readcoop.eu/wp-content/uploads/2022/03/Bychawa_1874.jpg" alt="" class="wp-image-27974" width="698" height="536" srcset="https://readcoop.eu/wp-content/uploads/2022/03/Bychawa_1874.jpg 931w, https://readcoop.eu/wp-content/uploads/2022/03/Bychawa_1874-300x230.jpg 300w, https://readcoop.eu/wp-content/uploads/2022/03/Bychawa_1874-768x590.jpg 768w, https://readcoop.eu/wp-content/uploads/2022/03/Bychawa_1874-16x12.jpg 16w, https://readcoop.eu/wp-content/uploads/2022/03/Bychawa_1874-600x461.jpg 600w" sizes="(max-width: 698px) 100vw, 698px" /></figure>

Russian Civil Records late XIX cent.

This model is based on the historical mining document &#8220;Schwazer Berglehenbuch&#8221; which was written by different scribes in the beginning of the 16th century (approx. 1515). As a result, the geographical core region of the document is the mining area Falkenstein in Schwaz (Tyrol, Austria). The manuscript is currently stored by the Tyrolean Regional Archive (Hs. 1587) Language used: Early New High German Mining Hub: https://transkribus.eu/r/mining-hub/#/staticDirectory



<figure class="wp-block-image size-large"><img decoding="async" width="1024" height="464" src="https://readcoop.eu/wp-content/uploads/2022/03/kurrent_1515_5-1024x464.jpg" alt="" class="wp-image-27961" srcset="https://readcoop.eu/wp-content/uploads/2022/03/kurrent_1515_5-1024x464.jpg 1024w, https://readcoop.eu/wp-content/uploads/2022/03/kurrent_1515_5-300x136.jpg 300w, https://readcoop.eu/wp-content/uploads/2022/03/kurrent_1515_5-768x348.jpg 768w, https://readcoop.eu/wp-content/uploads/2022/03/kurrent_1515_5-18x8.jpg 18w, https://readcoop.eu/wp-content/uploads/2022/03/kurrent_1515_5-600x272.jpg 600w, https://readcoop.eu/wp-content/uploads/2022/03/kurrent_1515_5.jpg 1281w" sizes="(max-width: 1024px) 100vw, 1024px" /></figure>

Kurrent_1515_ENHG_v3

This model is based on the historical mining document &#8220;Verleihbuch der Rattenberger Bergrichter&#8221; which was written by Hanns Hueber (Bergschreiber) between 1460 and 1463. The manuscript is currently stored by the Tyrolean Regional Archives (Hs. 37). Language used: Early New High German Mining Hub: https://transkribus.eu/r/mining-hub/#/staticDirectory 



<figure class="wp-block-image size-large"><img decoding="async" width="1024" height="486" src="https://readcoop.eu/wp-content/uploads/2022/03/bastarda_1460-1463_54-1024x486.jpg" alt="" class="wp-image-27955" srcset="https://readcoop.eu/wp-content/uploads/2022/03/bastarda_1460-1463_54-1024x486.jpg 1024w, https://readcoop.eu/wp-content/uploads/2022/03/bastarda_1460-1463_54-300x142.jpg 300w, https://readcoop.eu/wp-content/uploads/2022/03/bastarda_1460-1463_54-768x364.jpg 768w, https://readcoop.eu/wp-content/uploads/2022/03/bastarda_1460-1463_54-18x9.jpg 18w, https://readcoop.eu/wp-content/uploads/2022/03/bastarda_1460-1463_54-600x285.jpg 600w, https://readcoop.eu/wp-content/uploads/2022/03/bastarda_1460-1463_54.jpg 1514w" sizes="(max-width: 1024px) 100vw, 1024px" /></figure>

Bastarda_1460-1463_ENHG_v2

Based on the protocolls of various authorities in Stockholm 1727-1763. 



Find out more: https://transkribus.eu/r/stockholm-city-archives/#/



<figure class="wp-block-image size-full"><img decoding="async" width="978" height="405" src="https://readcoop.eu/wp-content/uploads/2022/01/stockholm.jpg" alt="" class="wp-image-27638" srcset="https://readcoop.eu/wp-content/uploads/2022/01/stockholm.jpg 978w, https://readcoop.eu/wp-content/uploads/2022/01/stockholm-300x124.jpg 300w, https://readcoop.eu/wp-content/uploads/2022/01/stockholm-768x318.jpg 768w, https://readcoop.eu/wp-content/uploads/2022/01/stockholm-18x7.jpg 18w, https://readcoop.eu/wp-content/uploads/2022/01/stockholm-600x248.jpg 600w, https://readcoop.eu/wp-content/uploads/2022/01/stockholm-145x60.jpg 145w, https://readcoop.eu/wp-content/uploads/2022/01/stockholm-217x90.jpg 217w" sizes="(max-width: 978px) 100vw, 978px" /></figure>

Stockholm Notaries 1700 2.1

Chronicling Novelty and the Amsterdam City Archives present a first general HTR+ model for handwritings from the Low Countries in the 17th Century. The model was trained on chronicles from nowadays Belgium and the Netherlands and on notarial deeds from Amsterdam. More than 1 890 000 words have been trained for this model an the CER is 6.7% for the Pylaia-version. 







<figure class="wp-block-image size-full"><img decoding="async" width="752" height="390" src="https://readcoop.eu/wp-content/uploads/2021/12/Snippet_Dutch_XVII_C.png" alt="" class="wp-image-27401" srcset="https://readcoop.eu/wp-content/uploads/2021/12/Snippet_Dutch_XVII_C.png 752w, https://readcoop.eu/wp-content/uploads/2021/12/Snippet_Dutch_XVII_C-300x156.png 300w, https://readcoop.eu/wp-content/uploads/2021/12/Snippet_Dutch_XVII_C-18x9.png 18w, https://readcoop.eu/wp-content/uploads/2021/12/Snippet_Dutch_XVII_C-600x311.png 600w, https://readcoop.eu/wp-content/uploads/2021/12/Snippet_Dutch_XVII_C-116x60.png 116w, https://readcoop.eu/wp-content/uploads/2021/12/Snippet_Dutch_XVII_C-174x90.png 174w" sizes="(max-width: 752px) 100vw, 752px" /></figure>

Dutch XVII-Century

Dutch_XVII_Century

This model includes Transkribus German Kurrent M2 and some sets containing german language Latin script. It works best for documents/collections with Kurrent/Sütterlin and Latin script mixed. For only Kurrent or only Latin script the dedicated models might work better, so please test and compare before use.
This is a generic model trained on a diverse dataset. Such models provide good results without the need for any extra training work. However, the best results can usually be achieved by training a special model for homogenous material, e. g. texts written by the same person or from a narrow historical period.
Image source: Frey, Hans ( vermutlich 1885-1886): Geologie, Vorlesung von Albert Heim. (ETH-Bibliothek Zürich, Hs 1290:1,&nbsp;<a target="_blank" href="https://doi.org/10.7891/e-manuscripta-666" rel="noreferrer noopener">https://doi.org/10.7891/e-manuscripta-666</a>&nbsp;/ Public Domain Mark)&nbsp;(bearbeitet)



<figure class="wp-block-image size-large"><img decoding="async" width="898" height="1024" src="https://readcoop.eu/wp-content/uploads/2021/12/19093-898x1024.jpg" alt="" class="wp-image-27341" srcset="https://readcoop.eu/wp-content/uploads/2021/12/19093-898x1024.jpg 898w, https://readcoop.eu/wp-content/uploads/2021/12/19093-263x300.jpg 263w, https://readcoop.eu/wp-content/uploads/2021/12/19093-768x875.jpg 768w, https://readcoop.eu/wp-content/uploads/2021/12/19093-1348x1536.jpg 1348w, https://readcoop.eu/wp-content/uploads/2021/12/19093-11x12.jpg 11w, https://readcoop.eu/wp-content/uploads/2021/12/19093-600x684.jpg 600w, https://readcoop.eu/wp-content/uploads/2021/12/19093-53x60.jpg 53w, https://readcoop.eu/wp-content/uploads/2021/12/19093-79x90.jpg 79w, https://readcoop.eu/wp-content/uploads/2021/12/19093.jpg 1723w" sizes="(max-width: 898px) 100vw, 898px" /></figure>

Transkribus German handwriting M1

This model can decipher a wide range of Kurrent handwriting in german from the 16th to the early 19th century. Especially in the earlier centuries this model might work better than the usual Transkribus Kurrent model.
This is a generic model trained on a diverse dataset. Such models provide good results without the need for any extra training work. However, the best results can usually be achieved by training a special model for homogenous material, e. g. texts written by the same person or from a narrow historical period.
Image source: Johannes Hennemann Reising (1596), Artificia alchimica (Östereichische Nationalbibliothek, Cod. 11450,&nbsp;<a target="_blank" href="http://data.onb.ac.at/rec/AC13947349" rel="noreferrer noopener">http://data.onb.ac.at/rec/AC13947349</a>).



<figure class="wp-block-image size-large"><img decoding="async" width="1024" height="613" src="https://readcoop.eu/wp-content/uploads/2021/12/00000018-1024x613.jpg" alt="" class="wp-image-27326" srcset="https://readcoop.eu/wp-content/uploads/2021/12/00000018-1024x613.jpg 1024w, https://readcoop.eu/wp-content/uploads/2021/12/00000018-300x180.jpg 300w, https://readcoop.eu/wp-content/uploads/2021/12/00000018-768x460.jpg 768w, https://readcoop.eu/wp-content/uploads/2021/12/00000018-18x12.jpg 18w, https://readcoop.eu/wp-content/uploads/2021/12/00000018-600x359.jpg 600w, https://readcoop.eu/wp-content/uploads/2021/12/00000018-100x60.jpg 100w, https://readcoop.eu/wp-content/uploads/2021/12/00000018-150x90.jpg 150w, https://readcoop.eu/wp-content/uploads/2021/12/00000018.jpg 1167w" sizes="(max-width: 1024px) 100vw, 1024px" /></figure>

Transkribus Early Kurrent M1

Model trained on 77 Danish-Norwegian and 23 German small blackletter prints (broadside ballads &#8211; &#8220;skillingsviser&#8221;) compiled in &#8220;Peder Rafns visebog&#8221;, København 1583-1634.



In the training data the different printing practices of the period are reflected. The letters i, j, u and v are transcribed diplomatic; i.e. as they appear. Ligatures β, ꜩ and œ are transcribed s, tz and oe. n/m with a macron is trancribed nn/mm, and the rotunda ꝛ is transcribed r.



<figure class="wp-block-image size-full"><img decoding="async" width="493" height="313" src="https://readcoop.eu/wp-content/uploads/2021/11/image.png" alt="" class="wp-image-27248" srcset="https://readcoop.eu/wp-content/uploads/2021/11/image.png 493w, https://readcoop.eu/wp-content/uploads/2021/11/image-300x190.png 300w, https://readcoop.eu/wp-content/uploads/2021/11/image-18x12.png 18w, https://readcoop.eu/wp-content/uploads/2021/11/image-95x60.png 95w, https://readcoop.eu/wp-content/uploads/2021/11/image-142x90.png 142w" sizes="(max-width: 493px) 100vw, 493px" /></figure>

Danish-Norwegian (and German) Print 17th century

NorFraktur_1600_PyLaia

HTR+ model created from ground truth data resulting from the transcription and manual segmentation of a sample of 169 pages of these old address books, in particular volumes 1898 and 1923. This model is valid for automatically transcribing the 1903 and 1913 volumes and any other document printed in two columns and using the Latin alphabet, especially in French. The choice of the sample is made by alphabetical criterion because it is the mode of organisation of the information in this document. The brackets present in the document have not been segmented. 118 pages for training and 51 pages for validation.



<figure class="wp-block-image size-large"><img decoding="async" width="1024" height="463" src="https://readcoop.eu/wp-content/uploads/2021/11/snippit-1024x463.png" alt="" class="wp-image-27047" srcset="https://readcoop.eu/wp-content/uploads/2021/11/snippit-1024x463.png 1024w, https://readcoop.eu/wp-content/uploads/2021/11/snippit-300x136.png 300w, https://readcoop.eu/wp-content/uploads/2021/11/snippit-768x347.png 768w, https://readcoop.eu/wp-content/uploads/2021/11/snippit-18x8.png 18w, https://readcoop.eu/wp-content/uploads/2021/11/snippit-600x271.png 600w, https://readcoop.eu/wp-content/uploads/2021/11/snippit-133x60.png 133w, https://readcoop.eu/wp-content/uploads/2021/11/snippit-199x90.png 199w, https://readcoop.eu/wp-content/uploads/2021/11/snippit.png 1027w" sizes="(max-width: 1024px) 100vw, 1024px" /></figure>

Annuaires_Propietaires_Adr_Paris_1898_1923

This model is a general model for Norwegian handwriting from approx. 1820 to 1940. The model is trained on transcribed letters and manuscripts from the collection at the National Library of Norway.



<figure class="wp-block-image size-full"><img decoding="async" width="935" height="386" src="https://readcoop.eu/wp-content/uploads/2021/11/unnamed-1.png" alt="" class="wp-image-27181" srcset="https://readcoop.eu/wp-content/uploads/2021/11/unnamed-1.png 935w, https://readcoop.eu/wp-content/uploads/2021/11/unnamed-1-300x124.png 300w, https://readcoop.eu/wp-content/uploads/2021/11/unnamed-1-768x317.png 768w, https://readcoop.eu/wp-content/uploads/2021/11/unnamed-1-18x7.png 18w, https://readcoop.eu/wp-content/uploads/2021/11/unnamed-1-600x248.png 600w, https://readcoop.eu/wp-content/uploads/2021/11/unnamed-1-145x60.png 145w, https://readcoop.eu/wp-content/uploads/2021/11/unnamed-1-218x90.png 218w" sizes="(max-width: 935px) 100vw, 935px" /></figure>

NorHand 1820-1940

This model is a combination of handwritten private letters from World War II Finland (Tampere University Folklife Archives) and handwritten military unit diaries of the Finnish army of World War II (The National Archives of Finland). The model was trained by DIGIKÄKI &#8211; Digital History and Handwritten Sources project in Tampere University and the National Archives of Finland. For more information on DIGIKÄKI project, see:&nbsp;<a href="https://www.tuni.fi/en/research/digikaki-digital-history-and-handwritten-sources" target="_blank" rel="noreferrer noopener">https://www.tuni.fi/en/research/digikaki-digital-history-and-handwritten-sources</a>



<figure class="wp-block-image size-large"><img decoding="async" width="1024" height="583" src="https://readcoop.eu/wp-content/uploads/2021/09/Example-letter-1024x583.jpg" alt="" class="wp-image-26656" srcset="https://readcoop.eu/wp-content/uploads/2021/09/Example-letter-1024x583.jpg 1024w, https://readcoop.eu/wp-content/uploads/2021/09/Example-letter-300x171.jpg 300w, https://readcoop.eu/wp-content/uploads/2021/09/Example-letter-768x438.jpg 768w, https://readcoop.eu/wp-content/uploads/2021/09/Example-letter-18x10.jpg 18w, https://readcoop.eu/wp-content/uploads/2021/09/Example-letter-600x342.jpg 600w, https://readcoop.eu/wp-content/uploads/2021/09/Example-letter-105x60.jpg 105w, https://readcoop.eu/wp-content/uploads/2021/09/Example-letter-158x90.jpg 158w, https://readcoop.eu/wp-content/uploads/2021/09/Example-letter.jpg 1111w" sizes="(max-width: 1024px) 100vw, 1024px" /></figure>

Finnish_World_War_II_letters_and_diaries

Sotapäiväkirjat ja sota-ajan kirjeet

This HTR-model is based on the personal handwriting of Charles V of Spain. It is based upon political-institutional texts/ a letter of instructions to his son. It is based upon the book: Rachael Ball and Geoffrey Parker (eds.), Cómo ser rey. Instrucciones del emperador Carlos V a su hijo Felipe. Mayo de 1543. Spanish-English bilingual edition; jointly published with the CSA and The Hispanic Society of America; 2014; ISBN 978-84-15245-45-2



Examples of the handwriting can be found at: https://doi.org/10.5281/zenodo.5381739



Please cite this HTR-model as: Rachael Ball, Geoffrey Parker, the Hispanic Society of America, C.A Romein, Transkribus Public HTR Model: Carlos V/ Charles V Cómo ser rey_1543



<figure class="wp-block-image size-full"><img decoding="async" width="724" height="319" src="https://readcoop.eu/wp-content/uploads/2021/09/sample.png" alt="" class="wp-image-26587" srcset="https://readcoop.eu/wp-content/uploads/2021/09/sample.png 724w, https://readcoop.eu/wp-content/uploads/2021/09/sample-300x132.png 300w, https://readcoop.eu/wp-content/uploads/2021/09/sample-18x8.png 18w, https://readcoop.eu/wp-content/uploads/2021/09/sample-600x264.png 600w, https://readcoop.eu/wp-content/uploads/2021/09/sample-136x60.png 136w, https://readcoop.eu/wp-content/uploads/2021/09/sample-204x90.png 204w" sizes="(max-width: 724px) 100vw, 724px" /></figure>

Carlos V/ Charles V

Charlos V/ Charles V

Extended multi-language Transkribus print model, including antiqua and blackletter prints, typewriter, computer print outs and decorative fonts.



This model includes more languages than the print 0.3 model. The CER in M1 is higher than in 0.3 which is due to a more varied validation set. For languages that were already included in print 0.3 the M1 model usually performs equally well or slightly better than 0.3.



Curated by the Transkribus team, this model is occasionally updated with community data for continuous improvement.


<div class="wp-block-image">
<figure class="aligncenter size-full is-resized"><img decoding="async" src="https://readcoop.eu/wp-content/uploads/2021/09/PyLaia-model_3.jpg" alt="" class="wp-image-26582" width="741" height="283" srcset="https://readcoop.eu/wp-content/uploads/2021/09/PyLaia-model_3.jpg 434w, https://readcoop.eu/wp-content/uploads/2021/09/PyLaia-model_3-300x115.jpg 300w, https://readcoop.eu/wp-content/uploads/2021/09/PyLaia-model_3-18x7.jpg 18w" sizes="(max-width: 741px) 100vw, 741px" /></figure></div>

Transkribus Print Multi-Language

Transkribus Print M1

Model trained by Álvaro Cuéllar for the project ETSO: Estilometría aplicada al Teatro del Siglo de Oro (<a href="http://etso.es">http://etso.es</a>). It is able to modernize the spelling to the current rules. It has been trained with theatrical manuscripts, mostly 17th century.



If you want more information on how to use this model or you want to collaborate somehow, please write to <a href="mailto:alvarocuellar1995@hotmail.com">alvarocuellar1995@hotmail.com</a>.



If you use this model, please cite:



-Cuéllar, Álvaro. «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. 2023. -Cuéllar, Álvaro. &#8220;Spanish Golden Age Manuscripts (Spelling Modernization) 1.0&#8221;. Transkribus. 2021. -Cuéllar, Álvaro and Germán Vega García-Luengos. ETSO: Estilometría aplicada al Teatro del Siglo de Oro. 2017-2023. <a href="https://etso.es">https://etso.es</a>.



<figure class="wp-block-image size-full"><img decoding="async" width="728" height="530" src="https://readcoop.eu/wp-content/uploads/2021/07/spanish_theater_2.png" alt="" class="wp-image-26121" srcset="https://readcoop.eu/wp-content/uploads/2021/07/spanish_theater_2.png 728w, https://readcoop.eu/wp-content/uploads/2021/07/spanish_theater_2-300x218.png 300w, https://readcoop.eu/wp-content/uploads/2021/07/spanish_theater_2-16x12.png 16w, https://readcoop.eu/wp-content/uploads/2021/07/spanish_theater_2-600x437.png 600w, https://readcoop.eu/wp-content/uploads/2021/07/spanish_theater_2-82x60.png 82w, https://readcoop.eu/wp-content/uploads/2021/07/spanish_theater_2-124x90.png 124w" sizes="(max-width: 728px) 100vw, 728px" /></figure>

Spanish Golden Age Manuscripts (Spelling Modernization) 1.0

Spanish Golden Age Manuscripts (Spelling Modernization) 1.0

Model trained by Álvaro Cuéllar for the project ETSO. Estilometría aplicada al Teatro del Siglo de Oro (<a href="http://etso.es/" target="_blank" rel="noreferrer noopener">http://etso.es</a>). It is able to modernize the spelling to the current rules. It is trained with theatre prints, mostly 17th&nbsp;and 18th&nbsp;centuries. If you want more information on how to use this model or you want to collaborate somehow, please write to&nbsp;<a href="mailto:alvaro.cuellar@uky.edu" target="_blank" rel="noreferrer noopener">alvaro.cuellar@uky.edu</a>&nbsp;or&nbsp;<a href="mailto:alvarocuellar1995@hotmail.com" target="_blank" rel="noreferrer noopener">alvarocuellar1995@hotmail.com</a>.&nbsp;If you use this model, please cite -Cuéllar, Álvaro. &#8220;Spanish Golden Age Theatre Prints (Spelling Modernization) 1.0&#8221;. Transkribus. 2021, and -Cuéllar, Álvaro and Germán Vega García-Luengos. ETSO. Estilometría aplicada al Teatro del Siglo de Oro. 2017-2021.&nbsp;<a href="http://etso.es/" target="_blank" rel="noreferrer noopener">etso.es</a>.



<figure class="wp-block-image"><img alt=""/></figure>

Spanish Golden Age Theatre Prints (Spelling Modernization) 1.0

Spanish Golden Age Prints (Spelling Modernization) 1.0

Model developed from the &#8220;cahiers&#8221; (notebooks) of the ordinances of the intendants of New France (kept at Bibliothèque et Archives nationales du Québec). Written between 1705 and 1760, these notebooks present a good writing regularity and a standardized legal vocabulary. The model is based on 315 pages of transcripts taken from each of the 44 “cahiers” and achieves a CER of 4.18% on the validation set.



Transcriptions used for this model were produced by various partners of the Nouvelle-France Numérique/Digital New France&nbsp;project&nbsp;(see&nbsp;<a href="http://nouvellefrancenumerique.info/" target="_blank" rel="noreferrer noopener">nouvellefrancenumerique.info</a>)



<figure class="wp-block-image size-full"><img decoding="async" width="657" height="578" src="https://readcoop.eu/wp-content/uploads/2021/07/french.png" alt="" class="wp-image-26114" srcset="https://readcoop.eu/wp-content/uploads/2021/07/french.png 657w, https://readcoop.eu/wp-content/uploads/2021/07/french-300x264.png 300w, https://readcoop.eu/wp-content/uploads/2021/07/french-14x12.png 14w, https://readcoop.eu/wp-content/uploads/2021/07/french-600x528.png 600w, https://readcoop.eu/wp-content/uploads/2021/07/french-68x60.png 68w, https://readcoop.eu/wp-content/uploads/2021/07/french-102x90.png 102w" sizes="(max-width: 657px) 100vw, 657px" /></figure>

Ordonnances des Intendants

This model is a broad combination of French handwritings from late 17th&#8211; and early 18th-century. It is based on more than 1,600 pages (296,403 words) of correspondence and registers of colonial administrators in New France. It achieves a CER of 4.12% on validation set and includes the main types of writing of the period:&nbsp;“ronde”, “bâtarde” and “coulée”. However, it is not well suited to the writings of notaries or court clerks, but can be used as a basic template to train a model more adapted to these types of documents.



<figure class="wp-block-image size-full"><img decoding="async" width="701" height="558" src="https://readcoop.eu/wp-content/uploads/2021/07/snippit.png" alt="" class="wp-image-26111" srcset="https://readcoop.eu/wp-content/uploads/2021/07/snippit.png 701w, https://readcoop.eu/wp-content/uploads/2021/07/snippit-300x239.png 300w, https://readcoop.eu/wp-content/uploads/2021/07/snippit-16x12.png 16w, https://readcoop.eu/wp-content/uploads/2021/07/snippit-600x478.png 600w, https://readcoop.eu/wp-content/uploads/2021/07/snippit-75x60.png 75w, https://readcoop.eu/wp-content/uploads/2021/07/snippit-113x90.png 113w" sizes="(max-width: 701px) 100vw, 701px" /></figure>



Transcriptions used for this model were produced by various partners of the Nouvelle-France Numérique/Digital New France&nbsp;project&nbsp;(see&nbsp;<a href="http://nouvellefrancenumerique.info/" target="_blank" rel="noreferrer noopener">nouvellefrancenumerique.info</a>)

New France 17th-18th centuries

New France (17th-18th Century)

This second version of a generic model for handwritten Russian (predominantly late 19th/early 20th century) was trained as part of the MultiHTR project (Freiburg/Germany, www.multihtr.uni-freiburg.de). It incorporates models trained by the Estonian State Archive and the Hamburg-based INEL project. Portions of the GT data have kindly been provided by the Prozhito project (Saint Petersburg) and the Ukraine RD of JewishGen, USA. In some of the GT transcriptions, pre-1918 letters have been represented faithfully, while in other GT transcriptions, they have been replaced with their modern equivalents. We expanded the first version of the Russian generic model by adding the Russian Civil Records model by IAJGS, USA as well as several additional sources from the Prozhito database. Moreover, we incorporated data from the HKR dataset (https://github.com/abdoelsayed2016/HKR_Dataset).







<figure class="wp-block-image size-large"><img decoding="async" width="772" height="653" src="https://readcoop.eu/wp-content/uploads/2021/07/Image-12.07.21-at-14.58.jpg" alt="" class="wp-image-25851" srcset="https://readcoop.eu/wp-content/uploads/2021/07/Image-12.07.21-at-14.58.jpg 772w, https://readcoop.eu/wp-content/uploads/2021/07/Image-12.07.21-at-14.58-300x254.jpg 300w, https://readcoop.eu/wp-content/uploads/2021/07/Image-12.07.21-at-14.58-768x650.jpg 768w, https://readcoop.eu/wp-content/uploads/2021/07/Image-12.07.21-at-14.58-14x12.jpg 14w, https://readcoop.eu/wp-content/uploads/2021/07/Image-12.07.21-at-14.58-600x508.jpg 600w, https://readcoop.eu/wp-content/uploads/2021/07/Image-12.07.21-at-14.58-71x60.jpg 71w, https://readcoop.eu/wp-content/uploads/2021/07/Image-12.07.21-at-14.58-106x90.jpg 106w" sizes="(max-width: 772px) 100vw, 772px" /></figure>

Russian Generic Handwriting 2

Russian generic handwriting 2

The &#8220;Admiraliteit Zeeland 1605-1609 compleet&#8221;-model could be described as:



Model based on the criminal verdicts 1605-1609 of the Admiralty of Zeeland Source: Zeeuws Archief, GIDS101   Gids Admiraliteit in Zeeland, (1574) 1586-1609 (1795) 46 &#8216;Criminele sententiën, by den rade van de admiraliteyt ghegheven&#8217;, 1584 jul. 25-1788 jul. 7



This model was created during a project of the Working Group Paleography in Zeeland (PaiZ) to make interesting texts available.



<div class="wp-block-image"><figure class="aligncenter size-large"><img decoding="async" width="1024" height="282" src="https://readcoop.eu/wp-content/uploads/2021/04/Admiraliteit-Zeeland-1024x282.gif" alt="" class="wp-image-23385" srcset="https://readcoop.eu/wp-content/uploads/2021/04/Admiraliteit-Zeeland-1024x282.gif 1024w, https://readcoop.eu/wp-content/uploads/2021/04/Admiraliteit-Zeeland-300x83.gif 300w, https://readcoop.eu/wp-content/uploads/2021/04/Admiraliteit-Zeeland-768x212.gif 768w, https://readcoop.eu/wp-content/uploads/2021/04/Admiraliteit-Zeeland-1536x423.gif 1536w, https://readcoop.eu/wp-content/uploads/2021/04/Admiraliteit-Zeeland-16x4.gif 16w, https://readcoop.eu/wp-content/uploads/2021/04/Admiraliteit-Zeeland-600x165.gif 600w, https://readcoop.eu/wp-content/uploads/2021/04/Admiraliteit-Zeeland-190x52.gif 190w, https://readcoop.eu/wp-content/uploads/2021/04/Admiraliteit-Zeeland-327x90.gif 327w" sizes="(max-width: 1024px) 100vw, 1024px" /></figure></div>

Dutch Admiralty of Zeeland 17th Century

The models were trained upon roughly 2500 pages (ca. 400k words) of&nbsp;handwritten&nbsp;transcriptions from the School of Scottish Studies Archives (<a href="https://www.ed.ac.uk/information-services/library-museum-gallery/cultural-heritage-collections/school-scottish-studies-archives" target="_blank" rel="noreferrer noopener">https://www.ed.ac.uk/information-services/library-museum-gallery/cultural-heritage-collections/school-scottish-studies-archives</a>). The transcriptions are from recordings of hero tales and international tales in Scottish Gaelic, made between 1949 and 1979. They were gathered from tradition bearers living in the Scottish Highlands and Islands. The majority of the transcriptions are from one hand (~80%) with the remainder split across approximately 10 hands. Thus, the generalisability of the models is limited at present. We hope to diversify the training data in the future by adding further hands.



<figure class="wp-block-image size-large"><img decoding="async" width="658" height="799" src="https://readcoop.eu/wp-content/uploads/2021/06/scottish_gaelic.png" alt="" class="wp-image-24834" srcset="https://readcoop.eu/wp-content/uploads/2021/06/scottish_gaelic.png 658w, https://readcoop.eu/wp-content/uploads/2021/06/scottish_gaelic-247x300.png 247w, https://readcoop.eu/wp-content/uploads/2021/06/scottish_gaelic-10x12.png 10w, https://readcoop.eu/wp-content/uploads/2021/06/scottish_gaelic-600x729.png 600w, https://readcoop.eu/wp-content/uploads/2021/06/scottish_gaelic-49x60.png 49w, https://readcoop.eu/wp-content/uploads/2021/06/scottish_gaelic-74x90.png 74w" sizes="(max-width: 658px) 100vw, 658px" /></figure>

Scottish Gaelic 1949-1979

gd_tale_large_mixed_v1

This general model is based on Spanish printed books in Gothic typeface and published between 15th-16th Century. Digitized images used to create the model come from different online repositories, mainly <a href="http://www.bne.es/es/Catalogos/BibliotecaDigitalHispanica/Inicio/index.html">Biblioteca Digital Hispánica</a>, the official repository of <a href="http://www.bne.es/es/Inicio/index.html">Biblioteca Nacional de España (BNE)</a>.



Around 20 pages per book were taken and several accomplished scholars participated in the transcription phase. The dataset of the current version corresponds to 20 documents, 25˙531 lines and 220˙904 words.



For further information concerning collaboration, general model’s features, dataset and transcription criteria, write to stefano.bazzaco.1@gmail.com or visit the following page: <a href="https://github.com/stefanobazzaco/HTR-model-SpanishGothic_XV-XVI_extended">https://github.com/stefanobazzaco/HTR-model-SpanishGothic_XV-XVI_extended</a>



The model will be freely available inside the Transkribus platform, in the public model section. Its private usage is Open Access but its handling for distribution/publication requires giving appropriate credits to authors and publisher.



Dataset DOI: <a href="https://doi.org/10.5281/zenodo.4888926">https://doi.org/10.5281/zenodo.4888926</a> (Creative Commons License CC BY-NC-ND 4.0)



To the creation of the HTR model took part the following projects:



<ul><li>Progetto Mambrino – Università degli Studi di Verona resp. Anna Bognolo, Stefano Neri <a href="https://www.mambrino.it/it">https://www.mambrino.it/it</a></li></ul>



<div style="height:20px" aria-hidden="true" class="wp-block-spacer"></div>



<ul><li>BIDISO (Biblioteca Digital Siglo de Oro) – Universidade da Coruña resp. Sagrario López Poza, Nieves Pena Sueiro <a href="https://www.bidiso.es/index.htm">https://www.bidiso.es/index.htm</a></li></ul>



<div style="height:20px" aria-hidden="true" class="wp-block-spacer"></div>



<ul><li>COMEDIC (Catálogo de obras medievales impresas en castellano) – Universidad de Zaragoza resp. María Jesús Lacarra <a href="https://comedic.unizar.es/">https://comedic.unizar.es/</a></li></ul>



<div style="height:20px" aria-hidden="true" class="wp-block-spacer"></div>



<ul><li>Progetto di Eccellenza: Le Digital Humanities applicate alle lingue e letterature straniere – Università degli Studi di Verona resp. Paolo Frassi <a href="https://www.dlls.univr.it/?ent=progetto&amp;id=5327">https://www.dlls.univr.it/?ent=progetto&amp;id=5327</a></li></ul>



<div style="height:50px" aria-hidden="true" class="wp-block-spacer"></div>



<figure class="wp-block-image size-large"><img decoding="async" width="1024" height="610" src="https://readcoop.eu/wp-content/uploads/2021/06/spanish_gothic-1024x610.jpg" alt="" class="wp-image-24615" srcset="https://readcoop.eu/wp-content/uploads/2021/06/spanish_gothic-1024x610.jpg 1024w, https://readcoop.eu/wp-content/uploads/2021/06/spanish_gothic-300x179.jpg 300w, https://readcoop.eu/wp-content/uploads/2021/06/spanish_gothic-768x457.jpg 768w, https://readcoop.eu/wp-content/uploads/2021/06/spanish_gothic-1536x915.jpg 1536w, https://readcoop.eu/wp-content/uploads/2021/06/spanish_gothic-2048x1219.jpg 2048w, https://readcoop.eu/wp-content/uploads/2021/06/spanish_gothic-16x10.jpg 16w, https://readcoop.eu/wp-content/uploads/2021/06/spanish_gothic-600x357.jpg 600w, https://readcoop.eu/wp-content/uploads/2021/06/spanish_gothic-101x60.jpg 101w, https://readcoop.eu/wp-content/uploads/2021/06/spanish_gothic-151x90.jpg 151w" sizes="(max-width: 1024px) 100vw, 1024px" /></figure>

Spanish Gothic 15th-16th Century

SpanishGothic_XV-XVI_extended_v1.2

This general model is based on different Spanish printed books in Redonda (Round) typeface and published between 16th-17th Century. Digitized images used to create the model come from different online repositories, mainly <a href="http://www.bne.es/es/Catalogos/BibliotecaDigitalHispanica/Inicio/index.html">Biblioteca Digital Hispánica</a>, the official repository of <a href="http://www.bne.es/es/Inicio/index.html">Biblioteca Nacional de España (BNE)</a>.



Around 20 pages per book were taken and several accomplished scholars participated in the transcription phase. The dataset of the current version corresponds to 21 documents, 15˙897 lines and 119˙856 words.



For further information concerning collaboration, general model’s features, dataset and transcription criteria, write to stefano.bazzaco.1@gmail.com or visit the following page: <a href="https://github.com/stefanobazzaco/HTR-model-SpanishRedonda_XVI-XVII_extended">https://github.com/stefanobazzaco/HTR-model-SpanishRedonda_XVI-XVII_extended</a>



The model will be freely available inside the Transkribus platform, in the public model section. Its private usage is Open Access but its handling for distribution/publication requires giving appropriate credits to authors and publisher.



Dataset DOI: <a href="https://doi.org/10.5281/zenodo.4889217">https://doi.org/10.5281/zenodo.4889217</a> (Creative Commons License CC BY-NC-ND 4.0)



To the creation of the HTR model took part the following projects:



<ul><li>Progetto Mambrino – Università degli Studi di Verona resp. Anna Bognolo, Stefano Neri <a href="https://www.mambrino.it/it">https://www.mambrino.it/it</a></li></ul>



<div style="height:20px" aria-hidden="true" class="wp-block-spacer"></div>



<ul><li>BIDISO (Biblioteca Digital Siglo de Oro) – Universidade da Coruña resp. Sagrario López Poza, Nieves Pena Sueiro <a href="https://www.bidiso.es/index.htm">https://www.bidiso.es/index.htm</a></li></ul>



<div style="height:20px" aria-hidden="true" class="wp-block-spacer"></div>



<ul><li>COMEDIC (Catálogo de obras medievales impresas en castellano) – Universidad de Zaragoza resp. María Jesús Lacarra <a href="https://comedic.unizar.es/">https://comedic.unizar.es/</a></li></ul>



<div style="height:20px" aria-hidden="true" class="wp-block-spacer"></div>



<ul><li>Progetto di Eccellenza: Le Digital Humanities applicate alle lingue e letterature straniere – Università degli Studi di Verona resp. Paolo Frassi <a href="https://www.dlls.univr.it/?ent=progetto&amp;id=5327">https://www.dlls.univr.it/?ent=progetto&amp;id=5327</a></li></ul>



<div style="height:60px" aria-hidden="true" class="wp-block-spacer"></div>



<figure class="wp-block-image size-large"><img decoding="async" width="1024" height="568" src="https://readcoop.eu/wp-content/uploads/2021/06/spanish_redonda-1024x568.jpg" alt="" class="wp-image-24612" srcset="https://readcoop.eu/wp-content/uploads/2021/06/spanish_redonda-1024x568.jpg 1024w, https://readcoop.eu/wp-content/uploads/2021/06/spanish_redonda-300x167.jpg 300w, https://readcoop.eu/wp-content/uploads/2021/06/spanish_redonda-768x426.jpg 768w, https://readcoop.eu/wp-content/uploads/2021/06/spanish_redonda-1536x853.jpg 1536w, https://readcoop.eu/wp-content/uploads/2021/06/spanish_redonda-2048x1137.jpg 2048w, https://readcoop.eu/wp-content/uploads/2021/06/spanish_redonda-16x9.jpg 16w, https://readcoop.eu/wp-content/uploads/2021/06/spanish_redonda-600x333.jpg 600w" sizes="(max-width: 1024px) 100vw, 1024px" /></figure>

Spanish Redonda (Round Script) 16th-17th Century

SpanishRedonda_sXVI-XVII_extended_v1.2

A general model for historical and modern Polish handwriting.



<figure class="wp-block-image size-large"><img decoding="async" width="1024" height="567" src="https://readcoop.eu/wp-content/uploads/2021/06/Polish-1024x567.jpg" alt="" class="wp-image-24577" srcset="https://readcoop.eu/wp-content/uploads/2021/06/Polish-1024x567.jpg 1024w, https://readcoop.eu/wp-content/uploads/2021/06/Polish-300x166.jpg 300w, https://readcoop.eu/wp-content/uploads/2021/06/Polish-768x425.jpg 768w, https://readcoop.eu/wp-content/uploads/2021/06/Polish-16x9.jpg 16w, https://readcoop.eu/wp-content/uploads/2021/06/Polish-600x332.jpg 600w, https://readcoop.eu/wp-content/uploads/2021/06/Polish-108x60.jpg 108w, https://readcoop.eu/wp-content/uploads/2021/06/Polish-163x90.jpg 163w, https://readcoop.eu/wp-content/uploads/2021/06/Polish.jpg 1201w" sizes="(max-width: 1024px) 100vw, 1024px" /><figcaption>Image Source: <a href="https://de.wikipedia.org/wiki/Pr%C3%A4sidentschaftswahl_in_Polen_1926_(31._Mai)">Wikipedia</a></figcaption></figure>

Polish &#8211; General Model

Transkribus Polish M2

A general model for historical and modern French handwriting.



<figure class="wp-block-image size-large"><img decoding="async" width="1024" height="780" src="https://readcoop.eu/wp-content/uploads/2021/06/French_example-1024x780.jpg" alt="" class="wp-image-24572" srcset="https://readcoop.eu/wp-content/uploads/2021/06/French_example-1024x780.jpg 1024w, https://readcoop.eu/wp-content/uploads/2021/06/French_example-300x229.jpg 300w, https://readcoop.eu/wp-content/uploads/2021/06/French_example-768x585.jpg 768w, https://readcoop.eu/wp-content/uploads/2021/06/French_example-16x12.jpg 16w, https://readcoop.eu/wp-content/uploads/2021/06/French_example-600x457.jpg 600w, https://readcoop.eu/wp-content/uploads/2021/06/French_example-79x60.jpg 79w, https://readcoop.eu/wp-content/uploads/2021/06/French_example-118x90.jpg 118w, https://readcoop.eu/wp-content/uploads/2021/06/French_example.jpg 1177w" sizes="(max-width: 1024px) 100vw, 1024px" /></figure>

French &#8211; General Model

Transkribus French Model 1

This model is trained on the Russian part of bilingual Evenki/Russian manuscripts by Konstantin Rychkov dated 1911-1913, following pre-reform Cyrillic orthography. These are texts of various genres (folklore, personal narratives, shamanistic rites) collected in different Evenki dialects and translated into Russian. The training set comprises 581 half-pages (59300 words) from several folders of the Rychkov archive. PyLaia engine, 72 epochs.



The model is provided by the INEL project (&#8220;Grammatical Descriptions, Corpora, and Language Technology for Indigenous Northern Eurasian Languages&#8221;), <a href="https://inel.corpora.uni-hamburg.de/">https://inel.corpora.uni-hamburg.de/</a>. The archive of Konstantin Rychkov is preserved at the Institute of Oriental Manuscripts RAS, <a href="http://www.orientalstudies.ru/">http://www.orientalstudies.ru/</a>. A lower-resolution version of selected manuscripts is accessible at <a href="http://www.orientalstudies.ru/rus/index.php?option=com_content&amp;task=view&amp;id=10326&amp;Itemid=149">http://www.orientalstudies.ru/rus/index.php?option=com_content&amp;task=view&amp;id=10326&amp;Itemid=149</a>


<div class="wp-block-image">
<figure class="aligncenter size-large"><img decoding="async" width="550" height="892" src="https://readcoop.eu/wp-content/uploads/2021/05/rychkov-F4-RUS-sample.jpg" alt="" class="wp-image-23867" srcset="https://readcoop.eu/wp-content/uploads/2021/05/rychkov-F4-RUS-sample.jpg 550w, https://readcoop.eu/wp-content/uploads/2021/05/rychkov-F4-RUS-sample-185x300.jpg 185w, https://readcoop.eu/wp-content/uploads/2021/05/rychkov-F4-RUS-sample-7x12.jpg 7w, https://readcoop.eu/wp-content/uploads/2021/05/rychkov-F4-RUS-sample-37x60.jpg 37w, https://readcoop.eu/wp-content/uploads/2021/05/rychkov-F4-RUS-sample-55x90.jpg 55w" sizes="(max-width: 550px) 100vw, 550px" /></figure></div>

<div class="wp-block-image">
<figure class="aligncenter size-large"><img decoding="async" width="636" height="719" src="https://readcoop.eu/wp-content/uploads/2021/05/rychkov-F5-RUS-sample.jpg" alt="" class="wp-image-23868" srcset="https://readcoop.eu/wp-content/uploads/2021/05/rychkov-F5-RUS-sample.jpg 636w, https://readcoop.eu/wp-content/uploads/2021/05/rychkov-F5-RUS-sample-265x300.jpg 265w, https://readcoop.eu/wp-content/uploads/2021/05/rychkov-F5-RUS-sample-11x12.jpg 11w, https://readcoop.eu/wp-content/uploads/2021/05/rychkov-F5-RUS-sample-600x678.jpg 600w, https://readcoop.eu/wp-content/uploads/2021/05/rychkov-F5-RUS-sample-53x60.jpg 53w, https://readcoop.eu/wp-content/uploads/2021/05/rychkov-F5-RUS-sample-80x90.jpg 80w" sizes="(max-width: 636px) 100vw, 636px" /></figure></div>

Russian Handwriting early 20th century

Russian-RychkovArchive-v1.0

The &#8220;NOSCEMUS General Model&#8221; is tailored towards recognizing Latin prints from the early modern period. Although the model is designed to recognize Latin prints set in Antiqua-based typefaces, it is also capable of recognizing passages in Greek and passages set in (German) Fraktur.



In creating the Ground Truth the following transcription guidlines were followed: &#8211; ligatures (e. g. Æ or æ, Œ or œ) and standard abbreviations (e.g. -que, -us, -tur, …mm…, …nn…) have been expanded &#8211; long s (ſ) was transcribed as a normal s &#8211; small caps were transcribed as majuscules &#8211; special characters and diacritics (e. g. &amp;, ë, ï or ę) were kept



The model was released by Stefan Zathammer and it is based on training data coming from the Digital Sourcebook of the NOSCEMUS project (<a href="https://transkribus.eu/r/noscemus/#/" target="_blank" rel="noreferrer noopener">https://transkribus.eu/r/noscemus/#</a>).



If you use the Noscemus model as a base model for your own model, or if your edition is based on a transcription made with the help of the Noscemus model, you are kindly requested to mention the Noscemus model.



The NOSCEMUS project (<a href="https://www.uibk.ac.at/projects/noscemus/" target="_blank" rel="noreferrer noopener">https://www.uibk.ac.at/projects/noscemus</a>) has received funding from the European Research Council (ERC) under the European Union’s Horizon 2020 research and innovation programme (grant agreement No. 741374).


<div class="wp-block-image">
<figure class="aligncenter size-large"><img decoding="async" width="647" height="691" src="https://readcoop.eu/wp-content/uploads/2021/05/beispielbild_noscemus_3.0.png" alt="" class="wp-image-23837" srcset="https://readcoop.eu/wp-content/uploads/2021/05/beispielbild_noscemus_3.0.png 647w, https://readcoop.eu/wp-content/uploads/2021/05/beispielbild_noscemus_3.0-281x300.png 281w, https://readcoop.eu/wp-content/uploads/2021/05/beispielbild_noscemus_3.0-11x12.png 11w, https://readcoop.eu/wp-content/uploads/2021/05/beispielbild_noscemus_3.0-600x641.png 600w, https://readcoop.eu/wp-content/uploads/2021/05/beispielbild_noscemus_3.0-56x60.png 56w, https://readcoop.eu/wp-content/uploads/2021/05/beispielbild_noscemus_3.0-84x90.png 84w" sizes="(max-width: 647px) 100vw, 647px" /></figure></div>

Printed Latin and Greek (also German, English, Italian) 15th-19th century | PyLaia

Noscemus GM 6

This model has been created within the scope of “Bulliot, Bibracte et moi”, winner of a French Ministry of Culture&#8217;s call for Innovative Digital Services. That is a “citizen science” project dedicated to transcript, the notebooks of Jacques-Gabriel Bulliot (1817-1902), who initiated the first excavations at the Bibracte archaeological site between 1867 and 1895 (https://bbm.hypotheses.org). About 147 000 words have been trained for this model and the CER is 8.20%. The CER is relatively high, however, because there are a lot of drawings and plans in which text and dimensions are oriented heterogeneously.



This model was trained in 2021 by Emmanuelle Perrin and Philippe Chassignet (project managers CNRS &#8211; Archéorient &#8211; UMR 5133, Lyon, France) with the participation of François Chagnot &#8211; Jean-Robert Grousset &#8211; Sylvie Grousset &#8211; Myriam Guillaumet &#8211; Isabelle Lagoutte &#8211; Marie-Hélène Lidec &#8211; Christine Pélardy Seignol &#8211; Bruno Ragon &#8211; Jean-Claude Seignol &#8211; Didier Vernet.


<div class="wp-block-image">
<figure class="aligncenter size-large"><img decoding="async" width="641" height="297" src="https://readcoop.eu/wp-content/uploads/2021/04/French-Bulliot.jpg" alt="" class="wp-image-23388" srcset="https://readcoop.eu/wp-content/uploads/2021/04/French-Bulliot.jpg 641w, https://readcoop.eu/wp-content/uploads/2021/04/French-Bulliot-300x139.jpg 300w, https://readcoop.eu/wp-content/uploads/2021/04/French-Bulliot-16x7.jpg 16w, https://readcoop.eu/wp-content/uploads/2021/04/French-Bulliot-600x278.jpg 600w, https://readcoop.eu/wp-content/uploads/2021/04/French-Bulliot-129x60.jpg 129w, https://readcoop.eu/wp-content/uploads/2021/04/French-Bulliot-194x90.jpg 194w" sizes="(max-width: 641px) 100vw, 641px" /></figure></div>

French Handwriting 19th century

BBM Bulliot French C19th handwritten 2021

This model is trained on about 105.000 words from the so-called “Verfachbücher” from the Tyrolean Pustertal valley and based on the public Model German_Kurrent_XVI-XVIII_M1. The ground truth material covers several rather typical handwriting styles for the given period, region, and type of administrative texts. The model was created within the projects “Reading in the Alps. Private book ownership in the Catholically dominated Central Alps 1750–1800” (FWF, P 29329) and “Living in the Alps” („Digitalisierung und Informationsextraktion für die Digital Humanities“ (DI4DH)) by Manuela Mayer, Michael Prokosch, Michael Span, Peter Andorfer and Martin Kirnbauer. 



More information about the projects behind you can find via these links: 



<a href="https://rita.acdh.oeaw.ac.at">https://rita.acdh.oeaw.ac.at</a>



<a href="https://rita-vfbr.acdh-dev.oeaw.ac.at/">https://rita-vfbr.acdh-dev.oeaw.ac.at</a>











<div class="wp-block-image"><figure class="aligncenter size-large"><img decoding="async" width="865" height="1024" src="https://readcoop.eu/wp-content/uploads/2021/04/Land-registers-Tyrol-cropped-865x1024.jpg" alt="" class="wp-image-22693" srcset="https://readcoop.eu/wp-content/uploads/2021/04/Land-registers-Tyrol-cropped-865x1024.jpg 865w, https://readcoop.eu/wp-content/uploads/2021/04/Land-registers-Tyrol-cropped-254x300.jpg 254w, https://readcoop.eu/wp-content/uploads/2021/04/Land-registers-Tyrol-cropped-768x909.jpg 768w, https://readcoop.eu/wp-content/uploads/2021/04/Land-registers-Tyrol-cropped-1298x1536.jpg 1298w, https://readcoop.eu/wp-content/uploads/2021/04/Land-registers-Tyrol-cropped-10x12.jpg 10w, https://readcoop.eu/wp-content/uploads/2021/04/Land-registers-Tyrol-cropped-600x710.jpg 600w, https://readcoop.eu/wp-content/uploads/2021/04/Land-registers-Tyrol-cropped-51x60.jpg 51w, https://readcoop.eu/wp-content/uploads/2021/04/Land-registers-Tyrol-cropped-76x90.jpg 76w, https://readcoop.eu/wp-content/uploads/2021/04/Land-registers-Tyrol-cropped.jpg 1343w" sizes="(max-width: 865px) 100vw, 865px" /></figure></div>



© SLA, Bozen, VB St. Michaelsburg 1800, Bl. 189r.

Land registers Tyrol 1750-1800

This model is based on the handwritten resolutions of the Dutch States-General (1576-1796). It has been created during the project &#8216;REPUBLIC&#8217;, which aims to provide an online searchable edition of the Reolutions. For more information regarding the background of the model and how to cite it, please have a look <a href="https://republic.huygens.knaw.nl/index.php/verhalen/">here</a>.



During the first phase, Huygens ING have created a ground truth corpus of around 1,000 pages, with which several HTR-models have been trained. The latest model, which Huygens ING proudly releases for common use, is performing at a staggering 3.10% Character Error Rate. With the help of volunteers Huygens ING aims at lowering this percentage even further during the next phase of the project. Because of the excellent HTR results they are confident that the ultimate goal, making this important source (re)searchable from all sorts of perspectives, will become reality.



<figure class="wp-block-image size-large"><img decoding="async" width="680" height="1005" src="https://readcoop.eu/wp-content/uploads/2021/03/proj_republic2.jpg" alt="" class="wp-image-21439" srcset="https://readcoop.eu/wp-content/uploads/2021/03/proj_republic2.jpg 680w, https://readcoop.eu/wp-content/uploads/2021/03/proj_republic2-203x300.jpg 203w, https://readcoop.eu/wp-content/uploads/2021/03/proj_republic2-8x12.jpg 8w, https://readcoop.eu/wp-content/uploads/2021/03/proj_republic2-600x887.jpg 600w" sizes="(max-width: 680px) 100vw, 680px" /><figcaption>A Resolution from 1598 © Project REPUBLIC</figcaption></figure>

REPUBLIC 7 &#8211; 17th and 18th century Dutch

Republic_7

This new public model for the recognition of medieval Latin has been released thanks to a collaboration between the Bentham Project of the University College London and the DEEDS (Documents of Early England Data Set) Project of the University of Toronto. What is special: Transkribus can now read the peculiar type of handwritten medieval Latin found in 12th- to 15th- century legal manuscripts. This is “peculiar” because legal Latin is full of idiosyncratic abbreviations and random word hyphenations. It has been created with material from the DEEDS Project at the University of Toronto, established by Professor Michael Gervers in 1975, and later expanded with material from the St Frideswide and Eynsham Abby cartularies, held by Christ Church College, Oxford. More than 140 000 words have been trained and the CER on the Validation Set is 1.71%.



Contact: Chris Riley: c.riley@ucl.ac.uk.


<div class="wp-block-image">
<figure class="aligncenter size-large"><img decoding="async" width="975" height="883" src="https://readcoop.eu/wp-content/uploads/2021/03/UCL-University-of-Toronto.png" alt="" class="wp-image-21346" srcset="https://readcoop.eu/wp-content/uploads/2021/03/UCL-University-of-Toronto.png 975w, https://readcoop.eu/wp-content/uploads/2021/03/UCL-University-of-Toronto-300x272.png 300w, https://readcoop.eu/wp-content/uploads/2021/03/UCL-University-of-Toronto-768x696.png 768w, https://readcoop.eu/wp-content/uploads/2021/03/UCL-University-of-Toronto-13x12.png 13w, https://readcoop.eu/wp-content/uploads/2021/03/UCL-University-of-Toronto-600x543.png 600w, https://readcoop.eu/wp-content/uploads/2021/03/UCL-University-of-Toronto-66x60.png 66w, https://readcoop.eu/wp-content/uploads/2021/03/UCL-University-of-Toronto-99x90.png 99w" sizes="(max-width: 975px) 100vw, 975px" /></figure></div>

Medieval Latin 13th-15th Century

UCL–University of Toronto #7

This model is based on Ground Truth from the manuscripts Cod. Vind. Slav. 3 (Breviary of Vid of Omišalj) and II. beramski brevijar. It can be used for transcribing different hands of 14th and 15th century Croatian Glagolitic handwriting. The model automatically transcribes into the Latin script and is capable of dealing with ligatures and resolving the most common abbreviations. It was trained at the Slavic Department Freiburg (Achim Rabus). The GT data was provided by Sanja Zubčić (Rijeka) and Jagoda and Guido Kappel (Vienna). About 171 000 words have been trained and the CER on the Validation Set is 5.60%. 


<div class="wp-block-image">
<figure class="aligncenter size-large"><img decoding="async" width="538" height="237" src="https://readcoop.eu/wp-content/uploads/2021/03/Handwritten-Glagolitic.jpg" alt="" class="wp-image-21185" srcset="https://readcoop.eu/wp-content/uploads/2021/03/Handwritten-Glagolitic.jpg 538w, https://readcoop.eu/wp-content/uploads/2021/03/Handwritten-Glagolitic-300x132.jpg 300w, https://readcoop.eu/wp-content/uploads/2021/03/Handwritten-Glagolitic-16x7.jpg 16w, https://readcoop.eu/wp-content/uploads/2021/03/Handwritten-Glagolitic-136x60.jpg 136w, https://readcoop.eu/wp-content/uploads/2021/03/Handwritten-Glagolitic-204x90.jpg 204w" sizes="(max-width: 538px) 100vw, 538px" /></figure></div>

Glagolitic handwritten 14th and 15th century

Handwritten Glagolitic

The model is the result of a collaboration between the University Library Tübingen and the Slavic Department Freiburg (Achim Rabus). Selected pages from different printings of the Tübingen-Urach tradition have been taken as Ground Truth. About 28 000 words have been trained and the CER on the Validation Set is 3.51%. 


<div class="wp-block-image">
<figure class="aligncenter size-large"><img decoding="async" width="603" height="294" src="https://readcoop.eu/wp-content/uploads/2021/03/Glagolitic-pritings.jpg" alt="" class="wp-image-21187" srcset="https://readcoop.eu/wp-content/uploads/2021/03/Glagolitic-pritings.jpg 603w, https://readcoop.eu/wp-content/uploads/2021/03/Glagolitic-pritings-300x146.jpg 300w, https://readcoop.eu/wp-content/uploads/2021/03/Glagolitic-pritings-16x8.jpg 16w, https://readcoop.eu/wp-content/uploads/2021/03/Glagolitic-pritings-600x293.jpg 600w, https://readcoop.eu/wp-content/uploads/2021/03/Glagolitic-pritings-123x60.jpg 123w, https://readcoop.eu/wp-content/uploads/2021/03/Glagolitic-pritings-185x90.jpg 185w" sizes="(max-width: 603px) 100vw, 603px" /></figure></div>

Public AI Models

Start unlocking the past with Transkribus