The DOI matches its Crossref record; 30 of the 30 references it cites were found in a registry or an index; none of the 11 sources its claims cite is checked for support yet.
A Dataset of Wolof Ajami Manuscripts for HTR and OCR
Open accessAbstract
We present the first ever dataset of manually segmented and transcribed Ajami manuscripts written in Wolof.The term Ajami refers to modified Arabic-script orthographies used to transcribe African languages.Handwritten text recognition (HTR) and optical character recognition (OCR) models for Arabic-script languages perform poorly on African languages written in Ajami orthographies because these languages are not represented in the pre-training data of the models.This leads to recognition models being unable to extract unique Arabic-script letters and ubiquitous diacritics used in African languages, and struggling to adapt to various calligraphy styles used across Africa.We release the following as an open-source dataset: an ALTO formatting of high-quality images of handwritten and printed, 20th-century Wolof manuscripts; manual segmentation (region and line); and manual transcriptions.We extend our contribution by evaluating several Arabic-script recognition models intended for historical manuscripts and find they produce character error rates (CER) of 61-81%.Transcriptions produced by the evaluated recognition models, as well as a keyboard to transcribe Wolof Ajami manuscripts, are released as well.The digitally transcribed text in the dataset can also be utilized for various natural language processing (NLP) and historical linguistic tasks.
Overview
CoChat Summary
TL;DR
The paper presents the first dataset of manually segmented and transcribed Ajami manuscripts written in Wolof, comprising 19 20th-century manuscripts (4,743 lines, 313 pages) in ALTO format plus a specialized keyboard. It evaluates OpenITI Arabic-script recognition models and reports high character error rates of 61–81%.
Key findings
- The dataset contains 19 20th-century Wolof Ajami manuscripts covering 4,743 lines and 313 pages; 9 prose and 10 poetry manuscripts; 8 handwritten and 11 printed; all in Maghrebi calligraphy; provenance Touba, Senegal.
- Evaluated OpenITI Arabic-script HTR/OCR models produce CERs of 61–81% on Wolof Ajami; MS Mellon Print performs best (CER 0.625) and Urdu Print worst (CER 0.822).
- Models struggle to recognize Arabic-script diacritics (tashkīl) and unique Wolof Ajami letters/diacritics absent from their pre-training languages; top three models better capture general line structure but omit diacritics and unique Wolof characters.
- The authors release the ALTO-formatted dataset, model transcriptions, and a Keyman keyboard for transcribing historical Wolof Ajami manuscripts as open source.
Methods
- Curated 19 20th-century Wolof Ajami manuscripts from Boston University’s Ajami collection (“The Four Languages” and “African Ajami Library”); dataset includes high-quality images, manual region and line segmentation, line order, and manual transcriptions in ALTO format.
- Annotation used eScriptorium; each manuscript was transcribed by a philologist with Wolof Ajami expertise and then independently reviewed for error correction.
- Baseline experiments evaluated OpenITI’s suite of Arabic-script recognition models for historical manuscripts (MS Mellon Print, MS Pretrained, Generalized, Mellon Ottoturk Print, Mellon Print, Pretrained Print, Urdu Print) using character error rate (CER).
Limitations
- The paper does not include a dedicated limitations section; it states that existing recognition models perform poorly due to lack of appropriate training data for Wolof Ajami orthography and the heavy use of diacritics and unique letters absent from pre-training languages.
- Evident from the text: the dataset is limited to 20th-century Wolof Ajami manuscripts from Touba, Senegal, all in Maghrebi calligraphy, sourced from Boston University’s Ajami collection (8 handwritten, 11 printed).
- Evident from the text: the manuscript count is inconsistent—Section 2 reports 19 manuscripts, while the Conclusion reports 17 manuscripts; baseline evaluation is limited to OpenITI’s suite of Arabic-script models.
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 · 7 Not checked · 33 Own results and interpretations
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
“As seen in Table 3, we observe high CERs when these existing recognition models for historical Arabic-script manuscripts attempt to transcribe Wolof Ajami manuscripts.”
§ A Dataset of Wolof Ajami Manuscripts for HTR and OCR > 3.1. Results and Analysis
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.
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
“Furthermore, the models are also unable to recognize unique letters and diacritics that exist in the Wolof Ajami orthography (Figure 2) and are not represented in the training data for these models.”
§ A Dataset of Wolof Ajami Manuscripts for HTR and OCR > 3.1. Results and Analysis
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.
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
“Conversely, much longer words are produced by the Mellon Ottoturk Print and Mellon Print models.”
§ A Dataset of Wolof Ajami Manuscripts for HTR and OCR > 3.1. Results and Analysis
plainstates an association- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: unclear — It is unclear whether the paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
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
“We qualitatively observe in many examples that the top three models better recognize the general structure of the ground truth lines.”
§ A Dataset of Wolof Ajami Manuscripts for HTR and OCR > 3.1. Results and Analysis
hedgedstates an association- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: unclear — It is unclear whether the paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
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
“we have created a Keyman 1 keyboard with specifications that reflect the character inventory needed to transcribe historical Wolof manuscripts.”
§ A Dataset of Wolof Ajami Manuscripts for HTR and OCR > 2. Dataset and Curation
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: unclear — It is unclear whether the paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Rests on 2 assumptions not shown here.
18 more claims
Create a free account to see all 40 claims and to verify the 2 references still open.
- It may state two separate claims as one. 13
- Probably not a substantial claim. 5
Self-check
Percentages, sums and reported differences re-computed from the paper’s own numbers.
Each contradiction between two passages listed here was confirmed by an independent check.
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
§ A Dataset of Wolof Ajami Manuscripts for HTR and OCR > 4. Conclusion
“We find that these existing models produce high character error rates of 61-81% when trying to transcribe Wolof Ajami text.”
§ A Dataset of Wolof Ajami Manuscripts for HTR and OCR > 4. Conclusion
“| Urdu Print | 0.822 |”
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
§ A Dataset of Wolof Ajami Manuscripts for HTR and OCR > 2. Dataset and Curation
“We have curated 19 20th-century Wolof Ajami manuscripts covering 4,743 lines and 313 pages in total.”
§ A Dataset of Wolof Ajami Manuscripts for HTR and OCR > 4. Conclusion
“Our opensource dataset contains high-quality images of 17 manuscripts, manual segmentation for the region and line level, and diplomatic transcriptions provided by experts.”
Citations
Cites
No Language Left Behind: Scaling Human-Centered Machine Translation
Nllb Team, Costa‐jussà, M. R., Cross, J., Çelebi, O., Elbayad, M., Heafield, K., Heffernan, K. S., Kalbassi, E., Lam, J., Licht, D. J., et al. (2022) · arXiv (Cornell University) · doi:10.48550/arxiv.2207.04672 · Cited by 362
MasakhaNER 2.0: Africa-centric Transfer Learning for Named Entity Recognition
Adelani, D. I., Neubig, G., Ruder, S., Rijhwani, S., Beukman, M., Palen-Michel, C., Lignos, C., Alabi, J. O., Muhammad, S. H., Nabende, P., et al. (2022) · Conference on Empirical Methods in Natural Language Processing (EMNLP) · doi:10.18653/v1/2022.emnlp-main.298 · Cited by 219
Self-citationAssessing the Impact of OCR Quality on Downstream NLP Tasks
van Strien, D., Beelen, K., Ardanuy, M. C., Hosseini, K., McGillivray, B. C., & Colavizza, G. (2020) · International Conference on Agents and Artificial Intelligence (ICAART) · doi:10.5220/0009169004840496 · Cited by 151
A Survey of OCR in Arabic Language: Applications, Techniques, and Challenges
Faizullah, S., Ayub, M. S., Hussain, S., & Khan, M. A. (2023) · Applied Sciences · doi:10.3390/app13074584 · Cited by 114
The Arabic Script in Africa: Understudied Literacy
Mumin, M. (2014) · BRILL eBooks · doi:10.1163/9789004256804_004 · Cited by 96
MasakhaPOS: Part-of-Speech Tagging for Typologically Diverse African languages
Dione, C. M. B., Adelani, D. I., Nabende, P., Alabi, J. O., Sindane, T., Buzaaba, H., Muhammad, S. H., Emezue, C. C., Ogayo, P., Aremu, A., et al. (2023) · Proceedings of the conference - Association for Computational Linguistics. Meeting · doi:10.18653/v1/2023.acl-long.609 · Cited by 87
Alleviating Digitization Errors in Named Entity Recognition for Historical Documents
Boroş, E., Hamdi, A. T. E., Pontes, E. L., Cabrera-Diego, L. A., Moreno, J. G., Sidère, N., & Doucet, A. (2020) · Conference on Computational Natural Language Learning · doi:10.18653/v1/2020.conll-1.35 · Cited by 72
Advancements and Challenges in Arabic Optical Character Recognition: A Comprehensive Survey
Kasem, M. S., Mahmoud, M., & Kang, H. (2025) · ACM Computing Surveys · doi:10.1145/3768150 · Cited by 58
Muslims beyond the Arab World: The Odyssey of Ajami and the Muridiyya
Ngom, F. (2016) · doi:10.1093/acprof:oso/9780190279868.001.0001 · Cited by 50
OCR Improves Machine Translation for Low-Resource Languages
Ignat, O., Maillard, J., Chaudhary, V., & Guzmán, F. (2022) · Findings of the Association for Computational Linguistics: ACL (Findings of ACL) · doi:10.18653/v1/2022.findings-acl.92 · Cited by 43
AfroLM: A Self-Active Learning-based Multilingual Pretrained Language Model for 23 African LanguagesPreprint
Dossou, B. F. P., Tonja, A. L., Yousuf, O., Osei, S., Oppong, A., Shode, I., Awoyomi, O. O., & Emezue, C. C. (2022) · arXiv (Cornell University) · doi:10.48550/arxiv.2211.03263 · Cited by 36
Self-citationAdelani, D. I., Alabi, J. O., Fan, A., Kreutzer, J., Shen, X., Reid, M., Ruiter, D., Klakow, D., Nabende, P., Chang, E., et al. (2022) · arXiv (Cornell University) · doi:10.48550/arxiv.2205.02022 · Cited by 32
Self-citationAfriQA: Cross-lingual Open-Retrieval Question Answering for African LanguagesPreprint
Ogundepo, O., Gwadabe, T., Rivera, C. E., Clark, J. H., Ruder, S., Adelani, D. I., Dossou, B. F. P., Abdou Aziz, D. I. O. P., Sikasote, C., Hacheme, G., et al. (2023) · arXiv (Cornell University) · doi:10.48550/arxiv.2305.06897 · Cited by 32
Amharic OCR: An End-to-End Learning
Belay, B. H., Habtegebrial, T., Meshesha, M., et al. (2020) · Applied Sciences · doi:10.3390/app10031117 · Cited by 23
Advances and Limitations in Open Source Arabic-Script OCR: A Case Study
Kiessling, B., Kurin, G., Miller, M. T., & Smail, K. (2021) · Digital Studies / Le champ numérique · doi:10.16995/dscn.8094 · Cited by 22
Computational modelling of an optical character recognition system for Yorùbá printed text images
Olalekan Joseph, O. N. I., & Asahiah, F. Ọ. (2020) · Scientific African · doi:10.1016/j.sciaf.2020.e00415 · Cited by 13
A Historical Handwritten Dataset for Ethiopic OCR with Baseline Models and Human-Level Performance
Belay, B. H., Guyon, I., Mengiste, T., Tilahun, B., Liwicki, M., Asfaw, T. T., & Égelé, R. (2024) · Lecture notes in computer science · doi:10.1007/978-3-031-70543-4_2 · Cited by 9
A Handwritten Text Recognition Dataset for Ajami Manuscripts in Fulfulde and Hausa
Yousuf, O., Aminu, A., Muhammad, M. S., Usman, B., Hashim, M. K., Nivre, J., Megyesi, B., & Høgel, C. (2026) · Lecture Notes in Computer Science · doi:10.1007/978-3-032-04627-7_36 · Cited by 1
Self-citationWhere Are We? Evaluating LLM Performance on African Languages
Adebara, I., Toyin, H. O., Ghebremichael, N. T., Elmadany, A., & Abdul-Mageed, M. (2025) · arXiv · doi:10.48550/arxiv.2502.19582
eScriptorium: An Open Source Platform for Historical Document Analysis
Kiessling, B., Tissot, R., Stokes, P., & Ezra, D. S. B. (2019) · 2019 International Conference on Document Analysis and Recognition Workshops (ICDARW) · doi:10.1109/icdarw.2019.10032
Optical character recognition errors and their effects on natural language processing
Lopresti, D. (2008) · Proceedings of the second workshop on Analytics for noisy unstructured text data · doi:10.1145/1390749.1390753
Arabic font recognition based on diacritics features
Lutf, M., You, X., Cheung, Y.-M., & Chen, C. L. P. (2014) · Pattern Recognition · doi:10.1016/j.patcog.2013.07.015
Digitizing the Textual Heritage of the Premodern Islamicate World: Principles and Plans
Miller, M. T., Romanov, M. G., & Savant, S. B. (2018) · International Journal of Middle East Studies · doi:10.1017/s0020743817000964
West African Manuscripts in Arabic and African Languages and Digital Preservation
Ngom, F. (2017) · Oxford Research Encyclopedia of African History · doi:10.1093/acrefore/9780190277734.013.123
ʿAjamī Literacies of Africa: The Hausa, Fula, Mandinka, and Wolof Traditions
Ngom, F., Rodima-Taylor, D., & Robinson, D. (2023) · Islamic Africa · doi:10.1163/21540993-20230002
Showalter, E. H. (2016) · Proceedings of the Eighth International Conference on Information and Communication Technologies and Development · doi:10.1145/2909609.2909637
MasakhaNER 2.0: Africa-centric Transfer Learning for Named Entity Recognition
Adelani, D. I., Neubig, G., Ruder, S., Rijhwani, S., Beukman, M., Palen-Michel, C., Lignos, C., Alabi, J. O., Muhammad, S. H., Nabende, P., et al. (2022) · arXiv · doi:10.48550/arxiv.2210.12391
Self-citationMasakhaPOS: Part-of-Speech Tagging for Typologically Diverse African Languages
Dione, C. M. B., Adelani, D., Nabende, P., Alabi, J., Sindane, T., Buzaaba, H., Muhammad, S. H., Emezue, C. C., Ogayo, P., Aremu, A., et al. (2023) · arXiv · doi:10.48550/arxiv.2305.13989
OCR Improves Machine Translation for Low-Resource Languages
Ignat, O., Maillard, J., Chaudhary, V., & Guzmán, F. (2022) · arXiv · doi:10.48550/arxiv.2202.13274
Advances and Limitations in Open Source Arabic-Script OCR: A Case Study
Kiessling, B., Kurin, G., Miller, M. T., & Smail, K. (2024) · arXiv · doi:10.48550/arxiv.2402.10943
Research works printed but not linked: 2
- Maxim Romanov, Matthew Thomas Miller, Sarah Bowen Savant, and Masoumeh Seydi. 2019. Open islamicate texts initiative: A machine-readable corpus of texts produced in the premodern islamicate world (poster). In Digital Humanities 2019 Conference Papers (9-12 July 2019). Utrecht University, 2019.
- Eddie Antonio Santos and Atticus Harrigan. 2020. Design and evaluation of a smartphone key- board for plains cree syllabics. In Proceedings of the 1st Joint Workshop on Spoken Language Technologies for Under-Resourced Languages (SLTU) and Collaboration and Computing for Under-Resourced Languages (CCURL), pages 88-96.
Cited by
Brisville-Fertin, O. (2026) · Journal of Open Humanities Data · doi:10.5334/johd.646
Metadata
- Authors
- Yousuf, O., Diagne, E. D., Høgel, C., Megyesi, B., & Nivre, J.
- Affiliations
- Uppsala University (SE)
- Lund University (SE)
- Venue
- Proceedings of the Language Resources and Evaluation Conference
- Published
- 2026-05
- Pages
- 3234-3239
- ISSN
- 2522-2686
- OpenAlex
- W7160193388
- Semantic Scholar
- 0e8ada12f59df227408c7472cdf31e2cc09015e3
- Cited by
- 0
- References
- 9
- FWCI
- 1.29
- Type
- conference paper
- Language
- en
- Open access
- bronze
- doi.org
- Locations
- Language Resources and Evaluation Conference (LREC) (publishedVersion)
- Funders
- Vetenskapsrådet (2022-06343)
- Riksbankens Jubileumsfond (M24-0028)
- 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 › Natural Language Processing Techniques
- Keywords
- Ajami script, Wolof language, handwriting recognition, optical character recognition, historical manuscripts, character error rate, diacritics, calligraphic style, transcription, low-resource languages
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 2026-11-16, under version 3 of the claim definition (model deepseek-v4.1-flash, prompt extract-v3).
A second model labelled each claim on 2026-11-16 (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.