CTC, Attention, and Hybrid Models for Arabic Handwritten Text Line Recognition: A Comparative Study | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Research Article CTC, Attention, and Hybrid Models for Arabic Handwritten Text Line Recognition: A Comparative Study Omar Arjafellah, Abdellah Yousfi, Azhar Hadmi This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-9295189/v1 This work is licensed under a CC BY 4.0 License Status: Under Revision Version 1 posted 11 You are reading this latest preprint version Abstract Arabic handwriting is prevalent in official records, education, and documentary heritage across many regions worldwide, and ongoing digitization effortsincreasingly depend on reliable transcription of offline handwritten documents.These challenges are particularly pronounced in the Arab world, where largecollections of handwritten Arabic documents remain under-digitized. However,offline Arabic handwriting recognition remains challenging, and progress is difficult to quantify because published results are often not directly comparabledue to inconsistent experimental settings. In this work, we introduce a unified and reproducible evaluation protocol and use it to systematically comparethree major sequence modeling approaches for offline Arabic handwriting recognition: models trained with Connectionist Temporal Classification, attention-basedencoder–decoder models, and a hybrid strategy that integrates both. We reportperformance using Character Error Rate and complement headline results withtargeted qualitative and quantitative error analysis to characterize recurring errorpatterns and approach-specific failure modes. Experimental results reveal consistent performance differences across the evaluated approaches, with our hybridmodel achieving the best overall accuracy at CER = 8.58%. To our knowledge,this work provides the first controlled and reproducible evaluation of a joint CTC–attention hybrid approach for Arabic offline handwriting recognition alongside strong reference baselines, supporting subsequent research and large-scaledigitization workflows globally. Arabic handwriting recognition Offline handwriting recognition Connectionist Temporal Classification (CTC) Attention-based models Hybrid model Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Figure 6 Full Text Additional Declarations No competing interests reported. Supplementary Files T2.png T5.png T4.png T6.png T7.png T1.png T8.png T3.png T9.png Cite Share Download PDF Status: Under Revision Version 1 posted Editorial decision: Revision requested 28 Apr, 2026 Reviews received at journal 28 Apr, 2026 Reviews received at journal 27 Apr, 2026 Reviewers agreed at journal 17 Apr, 2026 Reviewers agreed at journal 15 Apr, 2026 Reviews received at journal 15 Apr, 2026 Reviewers agreed at journal 15 Apr, 2026 Reviewers invited by journal 15 Apr, 2026 Editor assigned by journal 02 Apr, 2026 Submission checks completed at journal 02 Apr, 2026 First submitted to journal 01 Apr, 2026 You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-9295189","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":626391724,"identity":"fe65f9f3-e8da-42a4-a549-4f0bb9ef8939","order_by":0,"name":"Omar Arjafellah","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAABEklEQVRIie3PP0vDQBjH8V84iEv1NklR2rdwJav4RlyeEmiWxqVrhkyZSudKQd9CxDfwyMFtri4RTZdODgGhOGjxAl2EHu0omC83Hc/n/gBtbX84AcEeV+AeIKsDiU9gu0Kgc+g9WzLM9hF5qlc10tcjmQtmSl/i4vmJ8ZniSma7SXcxCucwExEYn5jMKinKa/KmBsmcdxNVUtg8SuBtqph8bQkpHGdIMieJP4ANib6RNdNGx/YQ5X1bcuck4xBeTkKZjv17rsnuKNHcUjhIdzGewJuRGBhf8XCmBzfliPS5CZJ7B5Fn8QOwpqhnxLKq17p/UkaPy/f0Irl1kCbxBUS/duxw4J7fdrl3oq2tre3/9gO96mTKNXmtVgAAAABJRU5ErkJggg==","orcid":"","institution":"ENSIAS","correspondingAuthor":true,"prefix":"","firstName":"Omar","middleName":"","lastName":"Arjafellah","suffix":""},{"id":626391726,"identity":"31c2ce44-d35d-445d-9a77-82d048ab7369","order_by":1,"name":"Abdellah Yousfi","email":"","orcid":"","institution":"Faculty of Law, Economics and Social Sciences","correspondingAuthor":false,"prefix":"","firstName":"Abdellah","middleName":"","lastName":"Yousfi","suffix":""},{"id":626391734,"identity":"3506b0ce-95b2-4e8a-93e3-b903ac076f23","order_by":2,"name":"Azhar Hadmi","email":"","orcid":"","institution":"Higher Institute of Audiovisual and Cinema Professions (ISMAC)","correspondingAuthor":false,"prefix":"","firstName":"Azhar","middleName":"","lastName":"Hadmi","suffix":""}],"badges":[],"createdAt":"2026-04-01 18:24:50","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-9295189/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-9295189/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":107429042,"identity":"77a0e12c-b858-4871-8b13-ee02656ad0ae","added_by":"auto","created_at":"2026-04-21 12:02:05","extension":"pdf","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":87478,"visible":true,"origin":"","legend":"","description":"","filename":"arabicscriptfigure.pdf","url":"https://assets-eu.researchsquare.com/files/rs-9295189/v1/50565bc6c059c9e910905972.pdf"},{"id":107868268,"identity":"de25c6d6-908e-4a03-85cc-7fd815ca120e","added_by":"auto","created_at":"2026-04-27 07:09:45","extension":"pdf","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":104879,"visible":true,"origin":"","legend":"","description":"","filename":"encoder.pdf","url":"https://assets-eu.researchsquare.com/files/rs-9295189/v1/1e8f67bc9eca7127dfb5828c.pdf"},{"id":107489725,"identity":"2c0005e0-a8c6-4ecc-be82-5824517749d6","added_by":"auto","created_at":"2026-04-22 02:48:51","extension":"pdf","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":83304,"visible":true,"origin":"","legend":"","description":"","filename":"CTC.pdf","url":"https://assets-eu.researchsquare.com/files/rs-9295189/v1/3b3f706c4e2c244474a6e8f6.pdf"},{"id":107429047,"identity":"61f4d69c-e79b-4652-9b7b-478dc423908b","added_by":"auto","created_at":"2026-04-21 12:02:05","extension":"pdf","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":51771,"visible":true,"origin":"","legend":"","description":"","filename":"attention.pdf","url":"https://assets-eu.researchsquare.com/files/rs-9295189/v1/1403b3e809667f2122c9bbe6.pdf"},{"id":107704392,"identity":"3c334160-b245-436e-b2d8-f3bfaba1ee4c","added_by":"auto","created_at":"2026-04-24 08:45:10","extension":"pdf","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":49022,"visible":true,"origin":"","legend":"","description":"","filename":"Hybrid.pdf","url":"https://assets-eu.researchsquare.com/files/rs-9295189/v1/1619ff6cc0854e196feffc01.pdf"},{"id":107704405,"identity":"801bfa60-129c-4763-a9b2-d82935f0ab11","added_by":"auto","created_at":"2026-04-24 08:45:14","extension":"pdf","order_by":6,"title":"Figure 6","display":"","copyAsset":false,"role":"figure","size":12835,"visible":true,"origin":"","legend":"","description":"","filename":"Lambda.pdf","url":"https://assets-eu.researchsquare.com/files/rs-9295189/v1/2b50fd96079025e5b5eff22a.pdf"},{"id":107871342,"identity":"9706d9bd-1570-4690-9a23-9ef394e95c99","added_by":"auto","created_at":"2026-04-27 07:48:28","extension":"pdf","order_by":1,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":858753,"visible":true,"origin":"","legend":"","description":"","filename":"SpringerNatureLaTeXTemplate.pdf","url":"https://assets-eu.researchsquare.com/files/rs-9295189/v1_covered_1528373f-a98d-4c55-8be7-afd8f31911cd.pdf"},{"id":107429040,"identity":"e3ec933a-fc50-41c8-a710-866f38347a09","added_by":"auto","created_at":"2026-04-21 12:02:05","extension":"png","order_by":0,"title":"","display":"","copyAsset":false,"role":"supplement","size":46845,"visible":true,"origin":"","legend":"","description":"","filename":"T2.png","url":"https://assets-eu.researchsquare.com/files/rs-9295189/v1/1f265194d8313aa57d4bc6f4.png"},{"id":107488521,"identity":"0b241adb-6a22-41f4-b658-b091b07c76b0","added_by":"auto","created_at":"2026-04-22 02:44:57","extension":"png","order_by":1,"title":"","display":"","copyAsset":false,"role":"supplement","size":60115,"visible":true,"origin":"","legend":"","description":"","filename":"T5.png","url":"https://assets-eu.researchsquare.com/files/rs-9295189/v1/f84a55197c77ff6c8d7cb7ed.png"},{"id":107429043,"identity":"0441ff2b-6295-49e6-9809-12eb22e851d4","added_by":"auto","created_at":"2026-04-21 12:02:05","extension":"png","order_by":2,"title":"","display":"","copyAsset":false,"role":"supplement","size":129851,"visible":true,"origin":"","legend":"","description":"","filename":"T4.png","url":"https://assets-eu.researchsquare.com/files/rs-9295189/v1/13ec51d85f7c2c390617e095.png"},{"id":107429044,"identity":"757dfc9a-2475-43ea-8177-17015056837a","added_by":"auto","created_at":"2026-04-21 12:02:05","extension":"png","order_by":3,"title":"","display":"","copyAsset":false,"role":"supplement","size":56116,"visible":true,"origin":"","legend":"","description":"","filename":"T6.png","url":"https://assets-eu.researchsquare.com/files/rs-9295189/v1/0a69a4e869f06c42d8afe9f9.png"},{"id":107487193,"identity":"6a7dd02e-4f5e-4768-93f0-5a71542fe1a6","added_by":"auto","created_at":"2026-04-22 02:40:00","extension":"png","order_by":4,"title":"","display":"","copyAsset":false,"role":"supplement","size":86763,"visible":true,"origin":"","legend":"","description":"","filename":"T7.png","url":"https://assets-eu.researchsquare.com/files/rs-9295189/v1/53be6acd7f0a923ad44b0e66.png"},{"id":107429050,"identity":"33d49c68-ab50-4fb7-a50f-5f1278d47d8a","added_by":"auto","created_at":"2026-04-21 12:02:05","extension":"png","order_by":5,"title":"","display":"","copyAsset":false,"role":"supplement","size":210428,"visible":true,"origin":"","legend":"","description":"","filename":"T1.png","url":"https://assets-eu.researchsquare.com/files/rs-9295189/v1/6915b9b07b3aed8b5bfbe84e.png"},{"id":107487186,"identity":"d16556b4-61a8-4c06-a9b4-8b2564bebd13","added_by":"auto","created_at":"2026-04-22 02:39:58","extension":"png","order_by":6,"title":"","display":"","copyAsset":false,"role":"supplement","size":87961,"visible":true,"origin":"","legend":"","description":"","filename":"T8.png","url":"https://assets-eu.researchsquare.com/files/rs-9295189/v1/fd6c43e339d85ceb0fd64b15.png"},{"id":107429052,"identity":"39831bb7-4d70-4704-a89e-815b0fdbb57f","added_by":"auto","created_at":"2026-04-21 12:02:06","extension":"png","order_by":7,"title":"","display":"","copyAsset":false,"role":"supplement","size":102875,"visible":true,"origin":"","legend":"","description":"","filename":"T3.png","url":"https://assets-eu.researchsquare.com/files/rs-9295189/v1/b5f94ddafa04bbcfe9f6979f.png"},{"id":107704390,"identity":"eb5c4f5f-f178-41a4-a543-7b4c3de11562","added_by":"auto","created_at":"2026-04-24 08:45:10","extension":"png","order_by":8,"title":"","display":"","copyAsset":false,"role":"supplement","size":68778,"visible":true,"origin":"","legend":"","description":"","filename":"T9.png","url":"https://assets-eu.researchsquare.com/files/rs-9295189/v1/7a20c95e3ed7ff728f13448d.png"}],"financialInterests":"No competing interests reported.","formattedTitle":"CTC, Attention, and Hybrid Models for Arabic Handwritten Text Line Recognition: A Comparative Study","fulltext":[],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":false,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":false,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":true,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":true,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"
[email protected]","identity":"journal-of-king-saud-university-computer-and-information-sciences","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"","sideBox":"Learn more about [Journal of King Saud University Computer and Information Sciences](https://link.springer.com/journal/44443)","snPcode":"44443","submissionUrl":"https://submission.springernature.com/new-submission/44443/3","title":"Journal of King Saud University Computer and Information Sciences","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"stoa","reportingPortfolio":"Springer Open","inReviewEnabled":true,"inReviewRevisionsEnabled":true},"keywords":"Arabic handwriting recognition, Offline handwriting recognition, Connectionist Temporal Classification (CTC), Attention-based models, Hybrid model","lastPublishedDoi":"10.21203/rs.3.rs-9295189/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-9295189/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"Arabic handwriting is prevalent in official records, education, and documentary heritage across many regions worldwide, and ongoing digitization effortsincreasingly depend on reliable transcription of offline handwritten documents.These challenges are particularly pronounced in the Arab world, where largecollections of handwritten Arabic documents remain under-digitized. However,offline Arabic handwriting recognition remains challenging, and progress is difficult to quantify because published results are often not directly comparabledue to inconsistent experimental settings. In this work, we introduce a unified and reproducible evaluation protocol and use it to systematically comparethree major sequence modeling approaches for offline Arabic handwriting recognition: models trained with Connectionist Temporal Classification, attention-basedencoder–decoder models, and a hybrid strategy that integrates both. We reportperformance using Character Error Rate and complement headline results withtargeted qualitative and quantitative error analysis to characterize recurring errorpatterns and approach-specific failure modes. Experimental results reveal consistent performance differences across the evaluated approaches, with our hybridmodel achieving the best overall accuracy at CER = 8.58%. To our knowledge,this work provides the first controlled and reproducible evaluation of a joint CTC–attention hybrid approach for Arabic offline handwriting recognition alongside strong reference baselines, supporting subsequent research and large-scaledigitization workflows globally.","manuscriptTitle":"CTC, Attention, and Hybrid Models for Arabic Handwritten Text Line Recognition: A Comparative Study","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2026-04-21 12:01:47","doi":"10.21203/rs.3.rs-9295189/v1","editorialEvents":[{"type":"communityComments","content":0},{"type":"decision","content":"Revision requested","date":"2026-04-28T18:06:18+00:00","index":"","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2026-04-28T17:56:41+00:00","index":"hide","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2026-04-27T18:27:37+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"241634353034204945022239309208053046828","date":"2026-04-17T14:01:54+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"137243491683806599992931910519912433721","date":"2026-04-15T09:18:12+00:00","index":"hide","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2026-04-15T08:58:01+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"223733668229907295113996554757680736820","date":"2026-04-15T08:47:51+00:00","index":"hide","fulltext":""},{"type":"reviewersInvited","content":"","date":"2026-04-15T08:31:15+00:00","index":"","fulltext":""},{"type":"editorAssigned","content":"","date":"2026-04-02T12:37:05+00:00","index":"","fulltext":""},{"type":"checksComplete","content":"","date":"2026-04-02T12:36:59+00:00","index":"","fulltext":""},{"type":"submitted","content":"Journal of King Saud University Computer and Information Sciences","date":"2026-04-01T18:15:10+00:00","index":"","fulltext":""}],"status":"published","journal":{"display":true,"email":"
[email protected]","identity":"journal-of-king-saud-university-computer-and-information-sciences","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"","sideBox":"Learn more about [Journal of King Saud University Computer and Information Sciences](https://link.springer.com/journal/44443)","snPcode":"44443","submissionUrl":"https://submission.springernature.com/new-submission/44443/3","title":"Journal of King Saud University Computer and Information Sciences","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"stoa","reportingPortfolio":"Springer Open","inReviewEnabled":true,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"88d4db6c-3d24-40cc-b59a-0cfccd5ccabd","owner":[],"postedDate":"April 21st, 2026","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"in-revision","subjectAreas":[],"tags":[],"updatedAt":"2026-04-28T18:09:26+00:00","versionOfRecord":[],"versionCreatedAt":"2026-04-21 12:01:47","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-9295189","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-9295189","identity":"rs-9295189","version":["v1"]},"buildId":"XKTyCvWXoU3ODBz1xrDgd","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}
Text is read by the "Ask this paper" AI Q&A widget below.
Extraction quality varies by source — PMC NXML preserves structure
cleanly, OA-HTML may include some navigation residue, and OA-PDF can
have broken hyphenation. The publisher copy
(via DOI)
is the canonical version.