The DOI matches its Crossref record; 52 of the 52 references it cites were found in a registry or an index; none of the 30 sources its claims cite is checked for support yet.
Unlocking the archives: Using large language models to transcribe handwritten historical documents
Open accessAbstract
This study demonstrates that Large Language Models (LLMs) can transcribe historical handwritten documents with significantly higher accuracy than specialized Handwritten Text Recognition (HTR) software, while being faster and more cost-effective. We introduce an open-source software tool called Transcription Pearl that leverages these capabilities to automatically transcribe and correct batches of handwritten documents using commercially available multimodal LLMs from OpenAI, Anthropic, and Google. In tests on a diverse corpus of 18th/19th century English language handwritten documents, LLMs achieved Character Error Rates (CER) of 5.7 to 7% and Word Error Rates (WER) of 8.9 to 15.9%, improvements of 14% and 32% respectively over specialized state-of-the-art HTR software like Transkribus. Most significantly, when LLMs were then used to correct those transcriptions as well as texts generated by conventional HTR software, they achieved near-human levels of accuracy, that is CERs as low as 1.8% and WERs of 3.5%. The LLMs also completed these tasks 50 times faster and at approximately 1/50th the cost of proprietary HTR programs. These results demonstrate that when LLMs are incorporated into software tools like Transcription Pearl, they provide an accessible, fast, and highly accurate method for mass transcription of historical handwritten documents, significantly streamlining the digitization process.
Overview
CoChat Summary
TL;DR
This paper tests commercial multimodal LLMs for transcribing handwritten historical documents and introduces an open-source tool, Transcription Pearl. It reports that LLMs outperform specialized HTR software out of the box and, when used for heterogeneous correction, approach non-expert human accuracy at much lower cost and higher speed.
Key findings
- On out-of-the-box transcription, Claude Sonnet-3.5 achieved strict CER 7.3% and WER 15.9%, and modified CER 5.7% and WER 8.9%; Transkribus Titan achieved strict CER 8.0% and WER 19.7%, while Transkribus PyLaia achieved strict CER 10.3% and WER 27.0%.
- LLM correction improved heterogeneous outputs: Claude Sonnet-3.5 correcting Gemini transcriptions achieved strict CER 5.7% and WER 13.8%, and modified CER 4.1% and WER 7.0%; self-correction was not statistically significant, with gpt-4o-06-08-2024 declining by 1–2% on strict and modified metrics.
- LLMs correcting Transkribus outputs reduced strict WER by 59% to an average of 9% and strict CER by 61% to around 4%; the paper reports Claude Sonnet-3.5 reduced Titan character-level errors by 68% to 1.8% and PyLaia word-level errors by 74% to 3.5%.
- LLMs completed a batch of 50 transcriptions in about 30 seconds versus about 22 minutes for Transkribus, roughly 50 times faster; correction cost $0.005–$0.01 per page, and automated LLM transcriptions cost about 1,500 times less than human transcription and roughly 50 times less than comparable automated HTR services.
Methods
- The authors built Transcription Pearl, an open-source Python GUI (3,907 lines) that calls commercial multimodal LLMs via API; tests used OpenAI gpt-4o-06-08-2024, Anthropic Claude Sonnet 3.5, and Google Gemini 1.5 Pro 002 with zero-shot prompts, temperature 0.0, and Top-P 1.0.
- The testing corpus comprised 50 pages (9,911 words) of English documents from 1761–1827 in 33 hands: 13 black-and-white microfilm scans, 26 medium-to-high-quality scans/photographs, and 11 iPhone or handheld photographs; all were 96 dpi, and ground truth was created from a Transkribus transcription manually corrected and proofed.
- Each test ran the 50 documents through Transkribus PyLaia/Titan and the LLM functions ten times (500 pages per model per test), averaged results, reported high/low scores and standard deviations, and computed strict and modified CER/WER with Python libraries while recording latency and cost.
Limitations
- Stated: the dataset is relatively small and focused solely on English-language 18th/19th-century documents; results may differ for other periods and languages; the dataset is not public; open-source models like Llama were not tested due to computing resources; and the tests used a simple zero-shot baseline, so further prompting may improve results.
- Stated: Gemini recitation false positives were possible, with one document flagged roughly 1 in 30 times; training-data pre-exposure could artificially deflate error rates, so the authors excluded internet-available documents.
- Evident from the text: only three commercial LLMs were tested; human error rates are cited from other studies rather than measured on the same corpus; and the ground truth was produced with Transkribus plus manual correction, which may shape the measured error rates.
Model-written summary. Check it against the paper before relying on it.
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
0 Valid · 0 Not valid · 27 Not checked · 84 Own results and interpretations
A result the paper states — it cites no source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“LLMs can be used to correct transcriptions generated by other LLMs as well as Transkribus to levels of accuracy that would normally require extensive fine-tuning, that is above 96%.”
§ Unlocking the Archives: Large Language Models Achieve State-of-the-Art Performance on the Transcription of Handwritten Historical Documents > Introduction
plainstates a cause- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Rests on an assumption not shown here.
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“While false positives are certainly possible (as we discuss below), we found that nearly every document from well known collections at McGill University, Harvard University Archives, and the William Clements Library at the University of Michigan we tried were rejected.”
§ Unlocking the Archives: Large Language Models Achieve State-of-the-Art Performance on the Transcription of Handwritten Historical Documents > Testing Dataset
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Rests on an assumption not shown here.
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“improvements of 14% and 32% respectively over specialized state-of-the-art HTR software like Transkribus.”
§ Unlocking the Archives: Large Language Models Achieve State-of-the-Art Performance on the Transcription of Handwritten Historical Documents > Abstract
plainstates a cause- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Rests on an assumption not shown here.
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“In our testing, we found that frontier model LLMs produce more accurate transcriptions of handwritten historical English language documents than state-of-the-art HTR models”
§ Unlocking the Archives: Large Language Models Achieve State-of-the-Art Performance on the Transcription of Handwritten Historical Documents > Performance on Out of the Box Transcription Tasks
plainstates a cause- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Rests on an assumption not shown here.
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“The PyLaia model transcribed the test data set with an average strict CER of 10.3% and WER of 27.0%”
§ Unlocking the Archives: Large Language Models Achieve State-of-the-Art Performance on the Transcription of Handwritten Historical Documents > Performance on Out of the Box Transcription Tasks
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Rests on an assumption not shown here.
56 more claims
Create a free account to see all 111 claims and to verify the 6 references still open.
- Probably not a substantial claim. 24
- It may state two separate claims as one. 33
Self-check
Percentages, sums and reported differences re-computed from the paper’s own numbers.
Rated minor: it would not touch a result — a value no result rests on, a p-value that stays on the same side of the significance level, or a conflict that is clerical, resolved elsewhere in the paper, or outside its own findings.
In the paper
“Transkribus took about 22 minutes to transcribe all 50 pages, or around 40 seconds per page.”
§ Unlocking the Archives: Large Language Models Achieve State-of-the-Art Performance on the Transcription of Handwritten Historical Documents > Performance on Out of the Box Transcription Tasks
Reported 40 · recomputed 23.588
Citations
Cites
Vaswani, A., Shazeer, N., Parmar, N., Uszkoreit, J., Jones, L., Gomez, A. N., Kaiser, L., & Polosukhin, I. (2017) · arXiv · doi:10.48550/arxiv.1706.03762 · Cited by 120,404
Gradient-based learning applied to document recognition
Lecun, Y., Bottou, L., Bengio, Y., & Haffner, P. (1998) · Proceedings of the IEEE · doi:10.1109/5.726791 · Cited by 58,805
Learning representations by back-propagating errors
Rumelhart, D. E., Hinton, G. E., & Williams, R. J. (1986) · Nature · doi:10.1038/323533a0 · Cited by 31,892
Language Models are Few-Shot Learners
Brown, T. B., Mann, B., Ryder, N., Subbiah, M., Kaplan, J., Dhariwal, P., Neelakantan, A., Shyam, P., Sastry, G., Askell, A., et al. (2020) · arXiv · doi:10.48550/arxiv.2005.14165 · Cited by 23,372
Scaling Laws for Neural Language Models
Kaplan, J., McCandlish, S., Henighan, T., Brown, T. B., Chess, B., Child, R., Gray, S., Radford, A., Wu, J., & Amodei, D. (2020) · arXiv · doi:10.48550/arxiv.2001.08361 · Cited by 9,550 · cited 1×
“The larger the number of weighted parameters in a model, the more robust its statistical representation of language and its meaning (Kaplan et al. 2020).”
Universal Language Model Fine-tuning for Text Classification
Howard, J., & Ruder, S. (2018) · Proceedings of the 56th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers) · doi:10.18653/v1/p18-1031 · Cited by 4,100
On the Opportunities and Risks of Foundation Models
Bommasani, R., Hudson, D. A., Adeli, E., Altman, R., Arora, S., von Arx, S., Bernstein, M. S., Bohg, J., Bosselut, A., Brunskill, E., et al. (2021) · arXiv · doi:10.48550/arxiv.2108.07258 · Cited by 2,676
Online and off-line handwriting recognition: a comprehensive survey
Plamondon, R., & Srihari, S. N. (2000) · IEEE Transactions on Pattern Analysis and Machine Intelligence · doi:10.1109/34.824821 · Cited by 2,540
A Novel Connectionist System for Unconstrained Handwriting Recognition
Graves, A., Liwicki, M., Fernandez, S., Bertolami, R., Horst Bunke, A., & Schmidhuber, J. (2009) · IEEE Transactions on Pattern Analysis and Machine Intelligence · doi:10.1109/tpami.2008.137 · Cited by 2,041
Handwritten Optical Character Recognition (OCR): A Comprehensive Systematic Literature Review (SLR)
Memon, J., Sami, M., Khan, R. A., & Uddin, M. (2020) · IEEE Access · doi:10.1109/access.2020.3012542 · Cited by 656
Focusing Attention: Towards Accurate Text Recognition in Natural Images
Cheng, Z., Bai, F., Xu, Y., Zheng, G., Pu, S., & Zhou, S. (2017) · IEEE/CVF International Conference on Computer Vision (ICCV) · doi:10.1109/iccv.2017.543 · Cited by 541
Nguyen (2022) · doi:10.1145/3453476 · Cited by 286
Joint Line Segmentation and Transcription for End-to-End Handwritten Paragraph RecognitionPreprint
Bluche, T. (2016) · arXiv (Cornell University) · doi:10.48550/arxiv.1604.08352 · Cited by 151
Start, Follow, Read: End-to-End Full-Page Handwriting Recognition
Wigington, C., Tensmeyer, C., Davis, B., Barrett, W. A., Price, B. L., & Cohen, S. D. (2018) · Lecture notes in computer science · doi:10.1007/978-3-030-01231-1_23 · Cited by 122
The Prompt Report: A Systematic Survey of Prompt Engineering Techniques
Schulhoff, S., Ilie, M., Balepur, N., Kahadze, K., Liu, A., Si, C., Li, Y., Gupta, A., Han, H., Schulhoff, S., et al. (2024) · arXiv · doi:10.48550/arxiv.2406.06608 · Cited by 118
Nockels, J., Gooding, P., Ames, S., & Terras, M. (2022) · Archival Science · doi:10.1007/s10502-022-09397-0 · Cited by 110
Kahle, P., Colutto, S., Hackl, G., & Muhlberger, G. (2017) · 2017 14th IAPR International Conference on Document Analysis and Recognition (ICDAR) · doi:10.1109/icdar.2017.307 · Cited by 108
Internal Consistency and Self-Feedback in Large Language Models: A SurveyPreprint
Liang, X., Song, S., Zheng, Z., Wang, H., Yu, Q., Li, X., Li, Y., Wang, Y., Wang, Z., Xiong, F., et al. (2024) · arXiv (Cornell University) · doi:10.48550/arxiv.2407.14507 · Cited by 97
Advancements and Challenges in Handwritten Text Recognition: A Comprehensive Survey
AlKendi, W., Gechter, F., Heyberger, L., & Guyeux, C. (2024) · Journal of Imaging · doi:10.3390/jimaging10010018 · Cited by 96
Tafti, A. P., Baghaie, A., Assefi, M., Arabnia, H. R., Yu, Z., & Peissig, P. L. (2016) · Lecture notes in computer science · doi:10.1007/978-3-319-50835-1_66 · Cited by 93
ICDAR 2019 Competition on Post-OCR Text Correction
Rigaud, C., Doucet, A., Coustaty, M., & Moreux, J.-P. (2019) · IAPR International Conference on Document Analysis and Recognition (ICDAR) · doi:10.1109/icdar.2019.00255 · Cited by 93
Pan, L., Saxon, M., Xu, W., Nathani, D., Wang, X., & Wang, W. Y. (2024) · Transactions of the Association for Computational Linguistics · doi:10.1162/tacl_a_00660 · Cited by 86
State of the Field: Digital History
Romein, C. A., Kemman, M., Birkholz, J. M., Baker, J., de Gruijter, M., Meroño-Peñuela, A., Ries, T., Ros, R., & Scagliola, S. (2020) · History · doi:10.1111/1468-229x.12969 · Cited by 83
ICDAR2017 Competition on Post-OCR Text Correction
Chiron, G., Doucet, A., Coustaty, M., & Moreux, J.-P. (2017) · IAPR International Conference on Document Analysis and Recognition (ICDAR) · doi:10.1109/icdar.2017.232 · Cited by 63
Amrhein, C., & Clematide, S. (2018) · LDV-Forum/Journal for language technology and computational linguistics · doi:10.21248/jlcl.33.2018.218 · Cited by 52
The implications of handwritten text recognition for accessing the past at scale
Nockels, J., Gooding, P., & Terras, M. (2024) · Journal of Documentation · doi:10.1108/jd-09-2023-0183 · Cited by 42
Comparing Human and Machine Errors in Conversational Speech Transcription
Stolcke, A., & Droppo, J. (2017) · Interspeech 2017 · doi:10.21437/interspeech.2017-1544 · Cited by 41
A comparative effectiveness study of eSource used for data capture for a clinical research registry
Nordo, A. H., Eisenstein, E. L., Hawley, J., Vadakkeveedu, S., Pressley, M., Pennock, J., & Sanderson, I. (2017) · International Journal of Medical Informatics · doi:10.1016/j.ijmedinf.2017.04.015 · Cited by 38
ICFHR2018 Competition on Automated Text Recognition on a READ Dataset
Strauß, T., Leifert, G., Labahn, R., Hodel, T., & Mühlberger, G. (2018) · International Conference on Frontiers in Handwriting Recognition · doi:10.1109/icfhr-2018.2018.00089 · Cited by 33
End-to-End page-Level assessment of handwritten text recognition
Vidal, E., Toselli, A. H., Ríos-Vila, A., & Calvo-Zaragoza, J. (2023) · Pattern Recognition · doi:10.1016/j.patcog.2023.109695 · Cited by 32
Transcription Error Rates in Retrospective Chart Reviews
Feng, J. E., Anoushiravani, A. A., Tesoriero, P. J., Ani, L., Meftah, M., Schwarzkopf, R., & Leucht, P. (2020) · Orthopedics · doi:10.3928/01477447-20200619-10 · Cited by 31
Early evidence of how LLMs outperform traditional systems on OCR/HTR tasks for historical recordsPreprint
Kim, S., Baudru, J., Ryckbosch, W., Bersini, H., & Ginis, V. (2025) · arXiv (Cornell University) · doi:10.48550/arxiv.2501.11623 · Cited by 27
Sánchez-DelaCruz, E., & Loeza-Mejía, C.-I. (2024) · Applied Intelligence · doi:10.1007/s10489-024-05487-x · Cited by 23
Handwriting Recognition in Historical Documents with Multimodal LLMPreprint
Li, L. (2024) · arXiv (Cornell University) · doi:10.48550/arxiv.2410.24034 · Cited by 23
A Study of The OCR Development History and Directions of Development
Wang, J. (2023) · Highlights in Science Engineering and Technology · doi:10.54097/bm665j77 · Cited by 22
Zhou, Y., Keuper, M., & Fritz, M. (2025) · Proceedings of the conference - Association for Computational Linguistics. Meeting · doi:10.18653/v1/2025.acl-long.1278 · Cited by 19
Capurro, C., Provatorova, V., & Kanoulas, E. (2023) · Heritage · doi:10.3390/heritage6120392 · Cited by 14
GPT-4 Doesn't Know It's Wrong: An Analysis of Iterative Prompting for Reasoning Problems
Stechly, K., Marquez, M., & Kambhampati, S. (2023) · arXiv · doi:10.48550/arxiv.2310.12397 · Cited by 11
Terras, M., Anzinger, B., Gooding, P., Mühlberger, G., Prien, M., Nockels, J., Romein, C. A., Stauder, A., & Stauder, F. (2025) · Open Research Europe · doi:10.12688/openreseurope.18747.2 · Cited by 11
Self-attention Networks for Non-recurrent Handwritten Text Recognition
d’Arce, R., Norton, T. J. T., Hannuna, S., & Cristianini, N. (2022) · Lecture notes in computer science · doi:10.1007/978-3-031-21648-0_27 · Cited by 10
Shooting the Archives: Document Digitization for Historical–Geographical Collaboration 1
Keeling, A., & Sandlos, J. (2011) · History Compass · doi:10.1111/j.1478-0542.2011.00771.x · Cited by 9
McLean, M. A., Roberts, D. A., & Gibbs, M. (2024) · Historical Methods A Journal of Quantitative and Interdisciplinary History · doi:10.1080/01615440.2024.2398455 · Cited by 8
Correzione dell'OCR per Corpus-assisted Discourse Studies: un caso di studio su vecchi quotidiani
Fante, D. D., & Nunzio, G. M. D. (2021) · Umanistica Digitale · doi:10.6092/issn.2532-8816/13689 · Cited by 6
Bergen, T. (2019) · Routledge eBooks · doi:10.4324/9781351142007 · Cited by 5
Interviewing Nookomis and Other Reflections: The Promise of Community Collaboration
Leddy, L. C. (2010) · Oral History Forum d'histoire orale · Cited by 4
Self-citationEditing Historical Documents: A Handbook of Practice
Stevens, M. E., & Burg, S. B. (1997) · Rowman & Littlefield Publishers eBooks · doi:10.5771/9780759117532 · Cited by 4
Preparing Big Manuscript Data for Hierarchical Clustering with Minimal HTR Training
Perdiki, E. (2023) · Journal of Data Mining & Digital Humanities · doi:10.46298/jdmdh.10419 · Cited by 2
From Digitization and Images to Text and Content: Transkribus as a Case Study
Prebor, G. (2023) · Proceedings of the Association for Information Science and Technology · doi:10.1002/pra2.958 · Cited by 1
Digital Humanities Techniques for Asylum Studies in the Archive
Nihart, N. (2022) · Carolina Digital Repository (University of North Carolina at Chapel Hill) · doi:10.17615/4hnn-kh38 · Cited by 1
Handwritten Text Recognition Error Rate Reduction in Historical Documents using Naive Transcribers
Christlein, V., Nicolaou, A., Schlauwitz, T., Späth, S., & Maier, A. (2018) · Gesellschaft für Informatik e.V. · doi:10.18420/infdh2018-13
Research works printed but not linked: 10
- Huang J. X. Chen S. Mishra H. S. Zheng A. W. Yu X. Song and D. Zhou. 2024. Large Language Models Cannot Self-Correct Reasoning Yet. ICLR 2024 Conference. https://openreview.net/pdf?id=IkmD3fKBPQ.
- Humphries, M. “History and Generative AI”. Teaching History 57 (3):4–9. https://search.informit.org/doi/10.3316/informit.322748974380724. · DOI
- Humphries, M., and L. Leddy. 2023. The future of history. Intersections 6 (3):21–2. https://cha-shc.ca/wp-content/uploads/2024/05/Intersections-6.3-online.pdf.
- Ó Raghallaigh, B. 2022. Handwritten text recognition (HTR) for Irish-Language Folklore. Proceedings of the 4th Celtic Language Technology Workshop. European Language Resource Association. Marseille, France, June 2022. 121–6. https://aclanthology.org/2022.cltw-1.17.
- Ren S. B. Tomlinson R. Black and A. Torrance. Reconciling the contrasting narratives on the environmental impact of large language models Nature: Scientific Reports https://www.nature.com/articles/s41598-024-76682-6.
- Sinclair, M., W. Lamb, and B. Alex. 2022. Handwriting Recognition for Scottish Gaelic. Proceedings of the Celtic Language Technology Workshop within LREC2022. Theodorus Fransen, William Lamb and Delyth Prys, eds. Paris: European Language Resources Association. Language Resources and Evaluatoin Conference, Paris, 20-2 June 2022. https://aclanthology.org/2022.cltw-1.pdf.
- 10.48550/arxiv.1708.08615 · DOI
- 10.12688/openreseurope.18747.1 · DOI
- Thomas A. R. Gaizauskas and H. Lu. 2024. Leveraging LLMs for post-OCR correction of historical newspapers. The Gale Review. https://review.gale.com/2024/09/03/using-large-language-models-for-post-ocr-correction/.
- 10.48550/arxiv.2408.13586 · DOI
Cited by
Multimodal LLMs for OCR, OCR Post-Correction, and Named Entity Recognition in Historical Documents
Greif, G., Griesshaber, N., & Greif, R. (2025) · arXiv · doi:10.48550/arxiv.2504.00414 · Cited by 49
Benchmarking large language models for handwritten text recognition
Crosilla, G., Klic, L., & Colavizza, G. (2025) · Journal of Documentation · doi:10.1108/jd-03-2025-0082 · Cited by 34
Large Language Models to make museum archive collections more accessible
Reusens, M., Adams, A., & Baesens, B. (2025) · AI & Society · doi:10.1007/s00146-025-02227-8 · Cited by 16
Aligned music notation and lyrics transcription
Fuentes-Martínez, E., Ríos-Vila, A., Martínez-Sevilla, J. C., Rizo, D., & Calvo-Zaragoza, J. (2025) · Pattern Recognition · doi:10.1016/j.patcog.2025.112094 · Cited by 10
Long-Tail Knowledge in Large Language Models: Taxonomy, Mechanisms, Interventions and ImplicationsPreprint
Badhe, S., Shah, D., & Kathrotia, N. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2602.16201 · Cited by 10
Metadata
- Authors
- Humphries, M., Leddy, L. C., Downton, Q., Legace, M., McConnell, J., Murray, I., & Spence, E.
- Affiliations
- Wilfrid Laurier University (CA)
- Corresponding author
- Humphries, Mark
- Venue
- Historical Methods: A Journal of Quantitative and Interdisciplinary History
- Publisher
- Informa UK Limited
- Published
- 2025-05-26
- Volume
- 58
- Issue
- 3
- Pages
- 175-193
- ISSN
- 0161-5440 (ISSN-L)
- 1940-1906
- OpenAlex
- W4410761886
- Cited by
- 72
- References
- 52
- Citations by year
- 2026: 50 · 2025: 21
- FWCI
- 16.76
- Type
- journal article
- Language
- en
- Open access
- green
- doi.org
- Locations
- Historical Methods A Journal of Quantitative and Interdisciplinary History (publishedVersion)
- SSRN Electronic Journal (submittedVersion)
- arXiv (Cornell University) (submittedVersion, cc-by)
- arXiv (Cornell University) (cc-by)
- Funders
- Insight Grant
- Social Sciences and Humanities Research Council of Canada; Canada Research Chairs
- Topics
- Physical Sciences › Computer Science › Computer Vision and Pattern Recognition › Handwritten Text Recognition Techniques
- Social Sciences › Arts and Humanities › Literature and Literary Theory › Digital Humanities and Scholarship
- Physical Sciences › Computer Science › Artificial Intelligence › Topic Modeling
- Keywords
- large language models, handwritten text transcription, historical handwritten documents, multimodal large language models, character error rate, word error rate, digitization of historical documents
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
A language model drew the claims from the paper on 2026-11-13, under version 3 of the claim definition (model deepseek-v4.1-flash, prompt extract-v3).
A second model labelled each claim on 2026-11-13 (model jev-1.13-20260917, definition revision gold-guideline-v0.2).
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
Data sources
Bibliographic data: Semantic Scholar, OpenAlex, Crossref.
Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).
Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.
