{
  "name": "Bayanat Labs Open",
  "url": "https://www.bayanatlabs.com/open",
  "publisher": "Bayanat Labs",
  "license": "https://creativecommons.org/publicdomain/zero/1.0/",
  "license_note": "This catalogue (metadata only) is dedicated to the public domain under CC0 1.0. Each release keeps its own licence, given per entry.",
  "updated": "2026-10-19",
  "takedown_email": "hello@bayanatlabs.com",
  "releases": [
    {
      "slug": "labelbench",
      "type": "benchmark",
      "status": "results-only",
      "title": "Bayanat Labs LabelBench",
      "title_ar": "LabelBench من Bayanat Labs",
      "summary": "LabelBench is a reproducible audit of whether AI models can label Arabic data, by taxonomy, task and dialect region. This release publishes aggregate results, charts and a technical report under CC BY 4.0 and the Python code under Apache-2.0; the source datasets are not redistributed, so the code downloads each one from its original location.",
      "summary_ar": "LabelBench تدقيق قابل لإعادة الإنتاج يختبر قدرة نماذج الذكاء الاصطناعي على توسيم البيانات العربية، بحسب التصنيف والمهمة والمنطقة اللهجية. ينشر هذا الإصدار نتائج مجمّعة ورسومًا بيانية وتقريرًا تقنيًا بترخيص CC BY 4.0، والشيفرة المصدرية بلغة Python بترخيص Apache-2.0. ولا يعيد توزيع مجموعات البيانات المصدرية، إذ تنزّل الشيفرة كل مجموعة من موقعها الأصلي.",
      "url": "https://www.bayanatlabs.com/open/labelbench",
      "url_ar": "https://www.bayanatlabs.com/ar/open/labelbench",
      "canonical_home": "https://www.bayanatlabs.com/research/labelbench",
      "licence": "CC-BY-4.0",
      "licence_url": "https://creativecommons.org/licenses/by/4.0/",
      "licence_label": "CC BY 4.0 (results, charts, report); Apache-2.0 (code)",
      "parts": [
        {
          "name": "Aggregate results (CSV and JSON) and checksums",
          "licence": "CC-BY-4.0",
          "licence_url": "https://creativecommons.org/licenses/by/4.0/",
          "in_release": true
        },
        {
          "name": "Charts and the v0.1 technical report (PDF)",
          "licence": "CC-BY-4.0",
          "licence_url": "https://creativecommons.org/licenses/by/4.0/",
          "in_release": true
        },
        {
          "name": "Code: the labelbench Python package",
          "licence": "Apache-2.0",
          "licence_url": "https://www.apache.org/licenses/LICENSE-2.0",
          "in_release": true
        },
        {
          "name": "Source datasets (MASSIVE 1.1 ar-SA, AraTox, IADD DART subset)",
          "licence": null,
          "licence_url": null,
          "in_release": false
        }
      ],
      "version": "v0.1",
      "date_published": "2026-07-07",
      "date_modified": "2026-10-14",
      "families": [
        "bench"
      ],
      "modality": "benchmark",
      "languages": [
        "ar"
      ],
      "key_numbers": {
        "results": 12,
        "tasks": 4,
        "source-datasets": 3
      },
      "doi": "10.5281/zenodo.23355784",
      "version_doi": "10.5281/zenodo.23355785",
      "github": "https://github.com/Bayanat-Labs/labelbench",
      "huggingface": null,
      "pypi": "https://pypi.org/project/bayanat-labelbench/",
      "kaggle": null,
      "downloads": [
        {
          "label": "Results (CSV)",
          "url": "https://www.bayanatlabs.com/research/labelbench-results-v0.1.csv",
          "kind": "file"
        },
        {
          "label": "Results (JSON)",
          "url": "https://www.bayanatlabs.com/research/labelbench-results-v0.1.json",
          "kind": "file"
        },
        {
          "label": "Checksums (SHA-256)",
          "url": "https://www.bayanatlabs.com/research/labelbench-release-v0.1.sha256",
          "kind": "file"
        },
        {
          "label": "Technical report (PDF)",
          "url": "https://www.bayanatlabs.com/research/labelbench-paper-v0.1.pdf",
          "kind": "file"
        },
        {
          "label": "Code on GitHub",
          "url": "https://github.com/Bayanat-Labs/labelbench",
          "kind": "repo"
        },
        {
          "label": "PyPI package",
          "url": "https://pypi.org/project/bayanat-labelbench/",
          "kind": "package"
        },
        {
          "label": "Zenodo archive (all versions)",
          "url": "https://doi.org/10.5281/zenodo.23355784",
          "kind": "doi"
        }
      ],
      "citation": [
        "Bayanat Labs Research (2026). Arabic Annotation Is Not Binary: LabelBench, a Reproducible Audit of Automated Data Labeling. Bayanat Labs, Technical report v0.1. https://www.bayanatlabs.com/research/labelbench",
        "Bayanat Labs Research (2026). LabelBench (v0.1.1) [Software]. Zenodo. https://doi.org/10.5281/zenodo.23355785"
      ],
      "author": "Bayanat Labs Research"
    },
    {
      "slug": "arabic-agent-reliability-lab",
      "type": "tool",
      "status": "beta",
      "title": "Bayanat Labs Arabic Agent Reliability Lab",
      "title_ar": "مختبر موثوقية الوكلاء بالعربية من Bayanat Labs",
      "summary": "The Arabic Agent Reliability Lab is a local harness that tests Arabic AI agents in a synthetic commerce sandbox and grades their tool calls and final state, not how finished the answer sounds. It is a methods beta: the code is Apache-2.0, and the open scenarios and the published evidence receipts are CC BY 4.0.",
      "summary_ar": "مختبر موثوقية الوكلاء بالعربية أداة تعمل محليًا لاختبار وكلاء الذكاء الاصطناعي العاملين بالعربية في بيئة تجارة اصطناعية، وتقيّم استدعاءاتهم للأدوات والحالة النهائية للنظام، لا مدى اكتمال الإجابة في ظاهرها. وهو نسخة تجريبية من المنهجية: الشيفرة بترخيص Apache-2.0، والسيناريوهات المفتوحة وإيصالات الأدلة المنشورة بترخيص CC BY 4.0.",
      "url": "https://www.bayanatlabs.com/open/arabic-agent-reliability-lab",
      "url_ar": "https://www.bayanatlabs.com/ar/open/arabic-agent-reliability-lab",
      "canonical_home": "https://www.bayanatlabs.com/research/arabic-agent-reliability",
      "licence": "Apache-2.0",
      "licence_url": "https://www.apache.org/licenses/LICENSE-2.0",
      "licence_label": "Apache-2.0 (code); CC BY 4.0 (open scenarios and receipts)",
      "parts": [
        {
          "name": "Harness code and documentation",
          "licence": "Apache-2.0",
          "licence_url": "https://www.apache.org/licenses/LICENSE-2.0",
          "in_release": true
        },
        {
          "name": "Open synthetic scenarios (scenarios/open/)",
          "licence": "CC-BY-4.0",
          "licence_url": "https://creativecommons.org/licenses/by/4.0/",
          "in_release": true
        },
        {
          "name": "Methods-beta evidence receipts v0.1 (receipts.json and checksum)",
          "licence": "CC-BY-4.0",
          "licence_url": "https://creativecommons.org/licenses/by/4.0/",
          "in_release": true
        },
        {
          "name": "Held-out evaluation material, raw traces and private prompts",
          "licence": null,
          "licence_url": null,
          "in_release": false
        }
      ],
      "version": "v0.2.0",
      "date_published": "2026-08-05",
      "date_modified": "2026-10-14",
      "families": [
        "agents",
        "bench"
      ],
      "modality": "tool",
      "languages": [
        "ar",
        "en"
      ],
      "key_numbers": {
        "workflow-families": 21,
        "repeats": 3,
        "runs": 4
      },
      "doi": "10.5281/zenodo.23355782",
      "version_doi": "10.5281/zenodo.23355783",
      "github": "https://github.com/Bayanat-Labs/arabic-agent-reliability-lab",
      "huggingface": null,
      "pypi": null,
      "kaggle": null,
      "downloads": [
        {
          "label": "Code on GitHub",
          "url": "https://github.com/Bayanat-Labs/arabic-agent-reliability-lab",
          "kind": "repo"
        },
        {
          "label": "Evidence receipts (JSON)",
          "url": "https://www.bayanatlabs.com/research/arabic-agent-reliability/v0.1/receipts.json",
          "kind": "file"
        },
        {
          "label": "Receipts checksum (SHA-256)",
          "url": "https://www.bayanatlabs.com/research/arabic-agent-reliability/v0.1/receipts.sha256",
          "kind": "file"
        },
        {
          "label": "Zenodo archive (all versions)",
          "url": "https://doi.org/10.5281/zenodo.23355782",
          "kind": "doi"
        }
      ],
      "citation": [
        "Bayanat Labs (2026). Arabic Agent Reliability Lab (v0.2.0) [Software]. Zenodo. https://doi.org/10.5281/zenodo.23355783"
      ],
      "author": "Bayanat Labs"
    },
    {
      "slug": "arabic-retrieval-starter-kit",
      "type": "dataset",
      "status": "live",
      "title": "Bayanat Labs Arabic Retrieval Starter Kit",
      "title_ar": "مجموعة بدء تقييم الاسترجاع العربي من Bayanat Labs",
      "summary": "The Arabic Retrieval Starter Kit is a small, fully synthetic Arabic and English fixture for testing retrieval and RAG pipelines. It covers Gulf and MSA questions, Arabic questions over English evidence, superseded policy versions and unanswerable questions, with graded TREC qrels and evidence-linked answers, under CC BY 4.0.",
      "summary_ar": "مجموعة بدء تقييم الاسترجاع العربي بيانات اختبار صغيرة واصطناعية بالكامل، بالعربية والإنجليزية، لاختبار أنظمة البحث والتوليد المعزز بالاسترجاع (RAG). تغطي أسئلة باللهجة الخليجية والفصحى، وأسئلة عربية أدلتها بالإنجليزية، ونسخًا قديمة من السياسات حلّت محلها نسخ أحدث، وأسئلة لا إجابة لها في المصادر، مع أحكام صلة متدرجة بصيغة TREC qrels وإجابات مرتبطة بالأدلة، بترخيص CC BY 4.0.",
      "url": "https://www.bayanatlabs.com/open/arabic-retrieval-starter-kit",
      "url_ar": "https://www.bayanatlabs.com/ar/open/arabic-retrieval-starter-kit",
      "canonical_home": "https://www.bayanatlabs.com/research/arabic-retrieval-starter-kit",
      "licence": "CC-BY-4.0",
      "licence_url": "https://creativecommons.org/licenses/by/4.0/",
      "licence_label": "CC BY 4.0",
      "parts": [
        {
          "name": "corpus.jsonl, queries.jsonl, qrels.txt, judgments.jsonl, answers.jsonl",
          "licence": "CC-BY-4.0",
          "licence_url": "https://creativecommons.org/licenses/by/4.0/",
          "in_release": true
        },
        {
          "name": "README.md, LICENSE.txt and SHA256SUMS.txt",
          "licence": "CC-BY-4.0",
          "licence_url": "https://creativecommons.org/licenses/by/4.0/",
          "in_release": true
        }
      ],
      "version": "v0.1",
      "date_published": "2026-10-05",
      "date_modified": "2026-10-14",
      "families": [
        "search"
      ],
      "modality": "text",
      "languages": [
        "ar",
        "en"
      ],
      "key_numbers": {
        "passages": 12,
        "queries": 12,
        "judgments": 30,
        "answers": 12
      },
      "doi": "10.5281/zenodo.23355780",
      "version_doi": "10.5281/zenodo.23355781",
      "github": "https://github.com/Bayanat-Labs/arabic-retrieval-starter-kit",
      "huggingface": null,
      "pypi": null,
      "kaggle": null,
      "downloads": [
        {
          "label": "corpus.jsonl",
          "url": "https://www.bayanatlabs.com/research/arabic-retrieval-starter-kit/v0.1/corpus.jsonl",
          "kind": "file"
        },
        {
          "label": "queries.jsonl",
          "url": "https://www.bayanatlabs.com/research/arabic-retrieval-starter-kit/v0.1/queries.jsonl",
          "kind": "file"
        },
        {
          "label": "qrels.txt (TREC)",
          "url": "https://www.bayanatlabs.com/research/arabic-retrieval-starter-kit/v0.1/qrels.txt",
          "kind": "file"
        },
        {
          "label": "judgments.jsonl",
          "url": "https://www.bayanatlabs.com/research/arabic-retrieval-starter-kit/v0.1/judgments.jsonl",
          "kind": "file"
        },
        {
          "label": "answers.jsonl",
          "url": "https://www.bayanatlabs.com/research/arabic-retrieval-starter-kit/v0.1/answers.jsonl",
          "kind": "file"
        },
        {
          "label": "README",
          "url": "https://www.bayanatlabs.com/research/arabic-retrieval-starter-kit/v0.1/README.md",
          "kind": "file"
        },
        {
          "label": "SHA256SUMS.txt",
          "url": "https://www.bayanatlabs.com/research/arabic-retrieval-starter-kit/v0.1/SHA256SUMS.txt",
          "kind": "file"
        },
        {
          "label": "All files on GitHub",
          "url": "https://github.com/Bayanat-Labs/arabic-retrieval-starter-kit",
          "kind": "repo"
        },
        {
          "label": "Zenodo archive (all versions)",
          "url": "https://doi.org/10.5281/zenodo.23355780",
          "kind": "doi"
        }
      ],
      "citation": [
        "Bayanat Labs (2026). Arabic Retrieval Starter Kit (v0.1) [Data set]. Zenodo. https://doi.org/10.5281/zenodo.23355781"
      ],
      "author": "Bayanat Labs"
    },
    {
      "slug": "arabic-annotation-guidelines-template",
      "type": "guide",
      "status": "live",
      "title": "Bayanat Labs Arabic Annotation Guidelines Template",
      "title_ar": "قالب إرشادات توسيم البيانات العربية من Bayanat Labs",
      "summary": "A free template for writing Arabic annotation guidelines, with copy-ready sections that cover dialect and register tags, diacritics, clitics, Arabizi, sarcasm, negation and religious formulas, plus decision trees and a pilot plan. Bayanat Labs wrote it and publishes it under CC BY 4.0, so you can adapt and share it with attribution.",
      "summary_ar": "قالب مجاني لكتابة إرشادات توسيم البيانات العربية، بأقسام جاهزة للنسخ تغطي وسوم اللهجة والسجل اللغوي، والتشكيل، والضمائر والحروف المتصلة، والعربيزي، والسخرية، والنفي، والعبارات الدينية، إضافة إلى أشجار قرار وخطة لمشروع تجريبي. كتبته Bayanat Labs وتنشره بترخيص CC BY 4.0، فيمكنك تكييفه ومشاركته مع نسبته إلى المصدر.",
      "url": "https://www.bayanatlabs.com/open/arabic-annotation-guidelines-template",
      "url_ar": "https://www.bayanatlabs.com/ar/open/arabic-annotation-guidelines-template",
      "canonical_home": "https://www.bayanatlabs.com/insights/arabic-annotation-guidelines-template",
      "licence": "CC-BY-4.0",
      "licence_url": "https://creativecommons.org/licenses/by/4.0/",
      "licence_label": "CC BY 4.0",
      "parts": [
        {
          "name": "Template text, examples and decision trees on the guide page",
          "licence": "CC-BY-4.0",
          "licence_url": "https://creativecommons.org/licenses/by/4.0/",
          "in_release": true
        }
      ],
      "version": null,
      "date_published": "2026-09-24",
      "date_modified": "2026-10-15",
      "families": [
        "knowledge"
      ],
      "modality": "report",
      "languages": [
        "en",
        "ar"
      ],
      "key_numbers": {
        "sections": 12
      },
      "doi": null,
      "version_doi": null,
      "github": null,
      "huggingface": null,
      "pypi": null,
      "kaggle": null,
      "downloads": [
        {
          "label": "Read and copy the template",
          "url": "https://www.bayanatlabs.com/insights/arabic-annotation-guidelines-template",
          "kind": "page"
        }
      ],
      "citation": [
        "Bayanat Labs Research (2026). Arabic Annotation Guidelines: A Free Template for Dialect, Diacritics and Arabizi. Bayanat Labs. https://www.bayanatlabs.com/insights/arabic-annotation-guidelines-template"
      ],
      "author": "Bayanat Labs Research"
    },
    {
      "slug": "open-policies",
      "type": "report",
      "status": "live",
      "title": "Bayanat Labs Open-Programme Policy and Operating Manual",
      "title_ar": "سياسة البرنامج المفتوح ودليل التشغيل من Bayanat Labs",
      "summary": "The rules Bayanat Labs follows when it chooses, licenses and publishes open Arabic data, as files anyone can read, check and reuse: the open-programme policy (selection rubric, release tiers, licence matrix), the operating manual with its gated stages and definition of done, and bl-score, a command-line tool that applies the rubric to a scoping memo. In English and Arabic. Documentation, not legal advice.",
      "summary_ar": "القواعد التي تتبعها Bayanat Labs في اختيار البيانات العربية المفتوحة وترخيصها ونشرها، في ملفات يستطيع أي شخص قراءتها والتحقق منها وإعادة استخدامها: سياسة البرنامج المفتوح (معايير الاختيار، ومستويات النشر، ومصفوفة التراخيص)، ودليل التشغيل بمراحله ذات البوابات وتعريف الإنجاز، وأداة bl-score التي تطبّق معايير الاختيار على مذكرة تحديد النطاق من سطر الأوامر. بالعربية والإنجليزية. توثيق، وليس استشارة قانونية.",
      "url": "https://www.bayanatlabs.com/open/open-policies",
      "url_ar": "https://www.bayanatlabs.com/ar/open/open-policies",
      "canonical_home": null,
      "licence": "CC-BY-4.0",
      "licence_url": "https://creativecommons.org/licenses/by/4.0/",
      "licence_label": "CC BY 4.0 (text); Apache-2.0 (code)",
      "parts": [
        {
          "name": "Open-programme policy: text, selection rubric, release tiers, licence matrix and schemas (policy/)",
          "licence": "CC-BY-4.0",
          "licence_url": "https://creativecommons.org/licenses/by/4.0/",
          "in_release": true
        },
        {
          "name": "Operating manual: public edition in English and Arabic, stages, definition of done and crawler rules (manual/)",
          "licence": "CC-BY-4.0",
          "licence_url": "https://creativecommons.org/licenses/by/4.0/",
          "in_release": true
        },
        {
          "name": "bl-score, validators, scripts and tests",
          "licence": "Apache-2.0",
          "licence_url": "https://www.apache.org/licenses/LICENSE-2.0",
          "in_release": true
        }
      ],
      "version": "v1.0.1",
      "date_published": "2026-10-15",
      "date_modified": "2026-10-15",
      "families": [
        "knowledge"
      ],
      "modality": "report",
      "languages": [
        "en",
        "ar"
      ],
      "key_numbers": {
        "stages": 13,
        "definition-of-done": 12
      },
      "doi": null,
      "version_doi": null,
      "github": null,
      "huggingface": null,
      "pypi": null,
      "kaggle": null,
      "downloads": [
        {
          "label": "Repository on GitHub",
          "url": "https://github.com/Bayanat-Labs/open-policies",
          "kind": "repo"
        },
        {
          "label": "Open-data policy (summary)",
          "url": "https://www.bayanatlabs.com/open/policy",
          "kind": "page"
        },
        {
          "label": "How we build open datasets (summary of the manual)",
          "url": "https://www.bayanatlabs.com/open/how-we-build",
          "kind": "page"
        }
      ],
      "citation": [
        "Bayanat Labs (2026). Bayanat Labs open policies, version 1.0.1. Bayanat Labs. https://github.com/Bayanat-Labs/open-policies"
      ],
      "author": "Bayanat Labs"
    },
    {
      "slug": "dataset-template",
      "type": "tool",
      "status": "beta",
      "title": "Bayanat Labs Dataset Template and Release Kit",
      "title_ar": "قالب مجموعات البيانات وحزمة أدوات الإصدار من Bayanat Labs",
      "summary": "The repository template that new Bayanat Labs open datasets are generated from, starting with this version. A generated repository carries a bilingual card, a licence on every record, a provenance ledger, a claims ledger, release files and a pipeline skeleton. The release kit adds a card linter (bl-card-lint), a release-gate checker (bl-gates), a manifest builder (bl-manifest) and a card renderer (bl-card-render). The standard and its card checklist are explained on the dataset standard page.",
      "summary_ar": "قالب المستودعات الذي تُنشأ منه مجموعات البيانات المفتوحة الجديدة من Bayanat Labs بدءًا من هذا الإصدار. ويحمل المستودع المُنشأ منه بطاقة ثنائية اللغة، وترخيصًا لكل سجل، وسجلًا لمصادر البيانات، وسجلًا للتصريحات، وملفات الإصدار، وهيكلًا أوليًا لخط المعالجة. وتضيف حزمة أدوات الإصدار مدققًا للبطاقة (bl-card-lint)، ومدققًا لبوابات الإصدار (bl-gates)، وأداة لإنشاء قائمة الملفات (bl-manifest)، وأداة لتعبئة البطاقة (bl-card-render). وتشرح صفحة معيار مجموعات البيانات المعيارَ وقائمة التحقق الخاصة بالبطاقات.",
      "url": "https://www.bayanatlabs.com/open/dataset-template",
      "url_ar": "https://www.bayanatlabs.com/ar/open/dataset-template",
      "canonical_home": null,
      "licence": "Apache-2.0",
      "licence_url": "https://www.apache.org/licenses/LICENSE-2.0",
      "licence_label": "Apache-2.0 (code); CC BY 4.0 (standard and docs)",
      "parts": [
        {
          "name": "Copier template, release kit, schemas and scripts",
          "licence": "Apache-2.0",
          "licence_url": "https://www.apache.org/licenses/LICENSE-2.0",
          "in_release": true
        },
        {
          "name": "Release standard, card checklist and documentation (docs/, standard/)",
          "licence": "CC-BY-4.0",
          "licence_url": "https://creativecommons.org/licenses/by/4.0/",
          "in_release": true
        }
      ],
      "version": "v0.1.0",
      "date_published": "2026-10-15",
      "date_modified": "2026-10-15",
      "families": [
        "knowledge"
      ],
      "modality": "tool",
      "languages": [
        "en",
        "ar"
      ],
      "key_numbers": {},
      "doi": null,
      "version_doi": null,
      "github": null,
      "huggingface": null,
      "pypi": null,
      "kaggle": null,
      "downloads": [
        {
          "label": "Template on GitHub",
          "url": "https://github.com/Bayanat-Labs/dataset-template",
          "kind": "repo"
        },
        {
          "label": "The standard and the card checklist",
          "url": "https://www.bayanatlabs.com/open/standard",
          "kind": "page"
        }
      ],
      "citation": [
        "Bayanat Labs (2026). Bayanat Labs dataset template and release standard, version 0.1.0. Bayanat Labs. https://github.com/Bayanat-Labs/dataset-template"
      ],
      "author": "Bayanat Labs"
    },
    {
      "slug": "arabic-normalize",
      "type": "tool",
      "status": "beta",
      "title": "Bayanat Labs Arabic Text Normalizer (arabic-normalize)",
      "title_ar": "أداة توحيد النص العربي من Bayanat Labs (arabic-normalize)",
      "summary": "arabic-normalize applies a named normalisation policy, such as bayanat-norm-v1, to Arabic text and reports exactly what changed: the normalised text, a change log with the position of every change, the lossy changes, the diacritization ratio and the SHA-256 of the policy. It comes as a Python library, a JavaScript package in the same repository and a free in-browser tool, with conformance fixtures that both implementations pass.",
      "summary_ar": "تطبّق مكتبة arabic-normalize على النص العربي سياسة توحيد مسمّاة، مثل bayanat-norm-v1، وتبيّن بالضبط ما غيّرته: النص الموحَّد، وسجل التغييرات مع موضع كل تغيير، والتغييرات الفاقدة للمعلومات، ونسبة التشكيل، وقيمة SHA-256 للسياسة. وتأتي مكتبةً بلغة Python، وحزمةً بلغة JavaScript في المستودع نفسه، وأداةً مجانية تعمل في المتصفح، مع حالات اختبار للمطابقة يجتازها التطبيقان كلاهما.",
      "url": "https://www.bayanatlabs.com/open/arabic-normalize",
      "url_ar": "https://www.bayanatlabs.com/ar/open/arabic-normalize",
      "canonical_home": null,
      "licence": "Apache-2.0",
      "licence_url": "https://www.apache.org/licenses/LICENSE-2.0",
      "licence_label": "Apache-2.0 (code); CC BY 4.0 (conformance fixtures)",
      "parts": [
        {
          "name": "Python library, CLI and JavaScript package",
          "licence": "Apache-2.0",
          "licence_url": "https://www.apache.org/licenses/LICENSE-2.0",
          "in_release": true
        },
        {
          "name": "Conformance fixtures (dev and acceptance splits)",
          "licence": "CC-BY-4.0",
          "licence_url": "https://creativecommons.org/licenses/by/4.0/",
          "in_release": true
        },
        {
          "name": "In-browser normalizer on Hugging Face Spaces",
          "licence": "Apache-2.0",
          "licence_url": "https://www.apache.org/licenses/LICENSE-2.0",
          "in_release": true
        }
      ],
      "version": "v0.1.0",
      "date_published": "2026-10-15",
      "date_modified": "2026-10-15",
      "families": [
        "bench",
        "voice"
      ],
      "modality": "tool",
      "languages": [
        "ar",
        "en"
      ],
      "key_numbers": {
        "fixtures": 80,
        "dev-fixtures": 64,
        "acceptance-fixtures": 16
      },
      "doi": null,
      "version_doi": null,
      "github": "https://github.com/Bayanat-Labs/arabic-normalize",
      "huggingface": "https://huggingface.co/datasets/Bayanat-Labs/arabic-normalization-conformance",
      "pypi": null,
      "kaggle": null,
      "downloads": [
        {
          "label": "Open the normalizer",
          "url": "https://www.bayanatlabs.com/tools/arabic-normalizer",
          "kind": "page"
        },
        {
          "label": "Code on GitHub",
          "url": "https://github.com/Bayanat-Labs/arabic-normalize",
          "kind": "repo"
        },
        {
          "label": "Conformance fixtures on Hugging Face",
          "url": "https://huggingface.co/datasets/Bayanat-Labs/arabic-normalization-conformance",
          "kind": "hf"
        },
        {
          "label": "In-browser tool on Hugging Face Spaces",
          "url": "https://huggingface.co/spaces/Bayanat-Labs/arabic-normalizer",
          "kind": "hf"
        }
      ],
      "citation": [
        "Bayanat Labs (2026). arabic-normalize: Arabic text normalisation with named, versioned policies, version 0.1.0. Bayanat Labs. https://github.com/Bayanat-Labs/arabic-normalize"
      ],
      "author": "Bayanat Labs"
    },
    {
      "slug": "consent-templates",
      "type": "guide",
      "status": "live",
      "title": "Bayanat Labs Contributor Consent Templates",
      "title_ar": "قوالب موافقة المساهمين من Bayanat Labs",
      "summary": "Six documents for collecting speech, writing, annotations and sign-language video from consenting adult contributors for open Arabic datasets: a consent form, a contributor agreement, a performer release for speech synthesis, a privacy notice, a cross-border transfer statement and an explicit biometric consent for sign-language video. Each comes in English and Arabic, as Markdown and Word files, written for Saudi Arabia's Personal Data Protection Law (PDPL). They are templates: adapt them and take your own legal advice before use.",
      "summary_ar": "ست وثائق لجمع الكلام المسجَّل والنصوص المكتوبة والتوسيمات ومقاطع الفيديو بلغة الإشارة من مساهمين بالغين يوافقون على ذلك، لبناء مجموعات بيانات عربية مفتوحة: نموذج موافقة، واتفاقية مساهمين، وإذن من المؤدّي باستخدام صوته في تركيب الكلام، وإشعار خصوصية، وبيان بنقل البيانات إلى خارج المملكة، وموافقة صريحة على جمع البيانات الحيوية في فيديو لغة الإشارة. وتأتي كل وثيقة بالعربية والإنجليزية، بصيغتي Markdown وWord، وقد كُتبت وفق نظام حماية البيانات الشخصية في السعودية. وهي قوالب: كيّفها لمؤسستك واطلب استشارة قانونية خاصة بها قبل استخدامها.",
      "url": "https://www.bayanatlabs.com/open/consent-templates",
      "url_ar": "https://www.bayanatlabs.com/ar/open/consent-templates",
      "canonical_home": null,
      "licence": "CC-BY-4.0",
      "licence_url": "https://creativecommons.org/licenses/by/4.0/",
      "licence_label": "CC BY 4.0",
      "parts": [
        {
          "name": "Six templates in English and Arabic (Markdown)",
          "licence": "CC-BY-4.0",
          "licence_url": "https://creativecommons.org/licenses/by/4.0/",
          "in_release": true
        },
        {
          "name": "Word (DOCX) copies of the same templates",
          "licence": "CC-BY-4.0",
          "licence_url": "https://creativecommons.org/licenses/by/4.0/",
          "in_release": true
        }
      ],
      "version": "v1.0",
      "date_published": "2026-10-15",
      "date_modified": "2026-10-15",
      "families": [
        "voice",
        "knowledge"
      ],
      "modality": "document",
      "languages": [
        "en",
        "ar"
      ],
      "key_numbers": {},
      "doi": null,
      "version_doi": null,
      "github": null,
      "huggingface": null,
      "pypi": null,
      "kaggle": null,
      "downloads": [
        {
          "label": "Templates on GitHub",
          "url": "https://github.com/Bayanat-Labs/consent-templates",
          "kind": "repo"
        }
      ],
      "citation": [
        "Bayanat Labs (2026). Bayanat Labs contributor consent templates, version 1.0. Bayanat Labs. https://github.com/Bayanat-Labs/consent-templates"
      ],
      "author": "Bayanat Labs"
    },
    {
      "slug": "arabicwer",
      "type": "tool",
      "status": "live",
      "title": "Bayanat Labs ArabicWER: Arabic WER and CER Scorer",
      "title_ar": "ArabicWER من Bayanat Labs: حساب معدل خطأ الكلمات والحروف للعربية",
      "summary": "ArabicWER (arabicwer) computes the word error rate (WER) and character error rate (CER) of Arabic speech recognition under six named, versioned normalisation policies side by side, from raw text to lenient orthography, so every score can name the policy it used, for example WER (arabicwer-1/lenient). It comes as a Python package and command line with a jiwer-style API, a specification with exact rules and a deterministic alignment, conformance fixtures with expected values under every policy, a free in-browser calculator, and a case study that scores two open Whisper models on FLEURS Egyptian Arabic under every policy.",
      "summary_ar": "تحسب ArabicWER (arabicwer) معدل خطأ الكلمات (WER) ومعدل خطأ الحروف (CER) لأنظمة التعرّف الآلي على الكلام العربي وفق ست سياسات توحيد مسمّاة ذات إصدارات ثابتة جنبًا إلى جنب، من النص الخام إلى الإملاء المتسامح، فتذكر كل نتيجة السياسة التي حُسبت بها، مثل WER (arabicwer-1/lenient). وتأتي حزمةً بلغة Python وأداةً لسطر الأوامر بواجهة تشبه jiwer، ومعها مواصفة بقواعد دقيقة وقاعدة محاذاة محددة، وحالات اختبار للمطابقة بقيمها المتوقعة في كل سياسة، وحاسبة مجانية تعمل في المتصفح، ودراسة حالة تقيّم نموذجين مفتوحين من Whisper على العربية المصرية في FLEURS بكل سياسة.",
      "url": "https://www.bayanatlabs.com/open/arabicwer",
      "url_ar": "https://www.bayanatlabs.com/ar/open/arabicwer",
      "canonical_home": null,
      "licence": "Apache-2.0",
      "licence_url": "https://www.apache.org/licenses/LICENSE-2.0",
      "licence_label": "Apache-2.0 (code); CC BY 4.0 (policies, specification, fixtures and case-study outputs)",
      "parts": [
        {
          "name": "Python package and command line",
          "licence": "Apache-2.0",
          "licence_url": "https://www.apache.org/licenses/LICENSE-2.0",
          "in_release": true
        },
        {
          "name": "Policies, specification and conformance fixtures",
          "licence": "CC-BY-4.0",
          "licence_url": "https://creativecommons.org/licenses/by/4.0/",
          "in_release": true
        },
        {
          "name": "Case study: model outputs, scores and report",
          "licence": "CC-BY-4.0",
          "licence_url": "https://creativecommons.org/licenses/by/4.0/",
          "in_release": true
        },
        {
          "name": "In-browser calculator on Hugging Face Spaces",
          "licence": "Apache-2.0",
          "licence_url": "https://www.apache.org/licenses/LICENSE-2.0",
          "in_release": true
        }
      ],
      "version": "v0.1.0",
      "date_published": "2026-10-15",
      "date_modified": "2026-10-15",
      "families": [
        "voice",
        "bench"
      ],
      "modality": "tool",
      "languages": [
        "ar",
        "en"
      ],
      "key_numbers": {
        "dev-fixtures": 120,
        "acceptance-fixtures": 40,
        "policies": 6,
        "casestudy-utterances": 100
      },
      "doi": null,
      "version_doi": null,
      "github": null,
      "huggingface": null,
      "pypi": null,
      "kaggle": null,
      "downloads": [
        {
          "label": "Open the calculator",
          "url": "https://www.bayanatlabs.com/tools/arabicwer",
          "kind": "page"
        },
        {
          "label": "Code and specification on GitHub",
          "url": "https://github.com/Bayanat-Labs/arabicwer",
          "kind": "repo"
        },
        {
          "label": "Fixtures and case study on Hugging Face",
          "url": "https://huggingface.co/datasets/Bayanat-Labs/arabicwer-fixtures",
          "kind": "hf"
        },
        {
          "label": "In-browser calculator on Hugging Face Spaces",
          "url": "https://huggingface.co/spaces/Bayanat-Labs/arabicwer",
          "kind": "hf"
        }
      ],
      "citation": [
        "Bayanat Labs (2026). arabicwer: Arabic WER and CER under named, versioned normalisation policies, version 0.1.0. Bayanat Labs. https://github.com/Bayanat-Labs/arabicwer"
      ],
      "author": "Bayanat Labs"
    },
    {
      "slug": "arabic-unicode-conformance",
      "type": "benchmark",
      "status": "live",
      "title": "Bayanat Labs Arabic Unicode Conformance Suite",
      "title_ar": "حزمة اختبار مطابقة يونيكود للنص العربي من Bayanat Labs",
      "summary": "Arabic text normalisation differs from tool to tool: one removes harakat and tatweel, another also folds أ إ آ to ا, a third turns each haraka into a space, so word error rates, deduplication counts and search indexes built with different normalisers are not comparable. This suite pins the behaviour down. Each test case gives an input string with its code points and bytes, and the exact output under six named policies: identity, NFC, NFKC and the three arabic-normalize policies raw-safe-v1, bayanat-norm-v1 and asr-eval-v1. A Python runner checks a normaliser against the cases, and a conformance matrix records how Python and ICU Unicode normalisation, arabic-normalize, CAMeL Tools, PyArabic and the Whisper text normaliser behave on them.",
      "summary_ar": "يختلف توحيد النص العربي من أداة إلى أخرى: فواحدة تحذف الحركات والتطويل، وأخرى توحّد أيضًا أ إ آ إلى ا، وثالثة تستبدل بكل حركة مسافة، فلا تعود معدلات خطأ الكلمات وأعداد التكرار وفهارس البحث المبنية بأدوات مختلفة قابلة للمقارنة. وتضبط هذه الحزمة السلوك بدقة: تعطي كل حالة اختبار نصًا مدخلًا بنقاط ترميزه وبايتاته، والناتج الدقيق وفق ست سياسات مسمّاة: identity وNFC وNFKC وسياسات arabic-normalize الثلاث raw-safe-v1 وbayanat-norm-v1 وasr-eval-v1. ويختبر برنامج تشغيل بلغة Python أي أداة توحيد مقابل الحالات، وتسجّل مصفوفة المطابقة سلوك توحيد Unicode في Python وICU، وarabic-normalize، وCAMeL Tools، وPyArabic، وأداة توحيد النص في Whisper عليها.",
      "url": "https://www.bayanatlabs.com/open/arabic-unicode-conformance",
      "url_ar": "https://www.bayanatlabs.com/ar/open/arabic-unicode-conformance",
      "canonical_home": null,
      "licence": "CC-BY-4.0",
      "licence_url": "https://creativecommons.org/licenses/by/4.0/",
      "licence_label": "CC BY 4.0 (cases, policy specifications, documentation and reports); Apache-2.0 (code)",
      "parts": [
        {
          "name": "Test cases and expected outputs",
          "licence": "CC-BY-4.0",
          "licence_url": "https://creativecommons.org/licenses/by/4.0/",
          "in_release": true
        },
        {
          "name": "Policy specifications and documentation (English and Arabic)",
          "licence": "CC-BY-4.0",
          "licence_url": "https://creativecommons.org/licenses/by/4.0/",
          "in_release": true
        },
        {
          "name": "Conformance matrix and observed outputs",
          "licence": "CC-BY-4.0",
          "licence_url": "https://creativecommons.org/licenses/by/4.0/",
          "in_release": true
        },
        {
          "name": "Runner and adapters (Python package)",
          "licence": "Apache-2.0",
          "licence_url": "https://www.apache.org/licenses/LICENSE-2.0",
          "in_release": true
        }
      ],
      "version": "v0.1.0",
      "date_published": "2026-10-15",
      "date_modified": "2026-10-15",
      "families": [
        "bench",
        "knowledge"
      ],
      "modality": "benchmark",
      "languages": [
        "ar",
        "en"
      ],
      "key_numbers": {
        "cases": 300,
        "categories": 18,
        "policies": 6,
        "expected-outputs": 1800,
        "implementations": 7,
        "pipelines": 20
      },
      "doi": null,
      "version_doi": null,
      "github": null,
      "huggingface": null,
      "pypi": null,
      "kaggle": null,
      "downloads": [
        {
          "label": "Dataset on Hugging Face",
          "url": "https://huggingface.co/datasets/Bayanat-Labs/arabic-unicode-conformance",
          "kind": "hf"
        },
        {
          "label": "Code and runner on GitHub",
          "url": "https://github.com/Bayanat-Labs/arabic-unicode-conformance",
          "kind": "repo"
        },
        {
          "label": "Normative policy text (bayanat-norm-v1)",
          "url": "https://github.com/Bayanat-Labs/arabic-unicode-conformance/blob/v0.1.0/docs/BAYANAT-NORM-V1.md",
          "kind": "repo"
        },
        {
          "label": "Arabic Text Normalizer (arabic-normalize)",
          "url": "https://www.bayanatlabs.com/open/arabic-normalize",
          "kind": "page"
        }
      ],
      "citation": [
        "Bayanat Labs (2026). Arabic Unicode Conformance Suite, version 0.1.0. Bayanat Labs. https://huggingface.co/datasets/Bayanat-Labs/arabic-unicode-conformance"
      ],
      "author": "Bayanat Labs"
    },
    {
      "slug": "gcc-format-validators",
      "type": "dataset",
      "status": "live",
      "title": "Bayanat Labs GCC Format Validators and Arabic ITN/TN Test Pairs",
      "title_ar": "مراجع الصيغ الخليجية واختبارات التحويل بين المنطوق والمكتوب من Bayanat Labs",
      "summary": "A speech recogniser may write an amount out in words where the transcript should show digits, and a text-to-speech front end has to read digits, currency symbols and dates aloud. This release gives speech, document and fintech teams test pairs for both directions, inverse text normalisation (ITN) and text normalisation (TN), in Modern Standard Arabic, Saudi and Gulf Arabic and Egyptian Arabic, with Western and Arabic-Indic digits: cardinals, ordinals, money, Hijri and Gregorian dates, times, phone numbers, identity numbers, IBANs, percentages and measures. With them come reference tables and validators for the formats the pairs contain: Saudi, Emirati and Egyptian identity numbers, Kuwaiti and Bahraini identity numbers (structure only), IBAN structures and phone numbering plans of Arab League states, Saudi plate letters, an Umm al-Qura Hijri to Gregorian table cross-checked against ICU, and Arabic number and currency words. All identifiers, IBANs and phone numbers are generated values that pass the format checks; they are not real personal data.",
      "summary_ar": "قد يكتب نظام التعرّف على الكلام مبلغًا بالحروف حيث يجب أن يظهر بالأرقام في النص، وعلى واجهة تحويل النص إلى كلام أن تنطق الأرقام ورموز العملات والتواريخ. ويقدّم هذا الإصدار لفرق الكلام والمستندات والتقنية المالية أزواج اختبار في الاتجاهين، من المنطوق إلى المكتوب (ITN) ومن المكتوب إلى المنطوق (TN)، بالعربية الفصحى والسعودية الخليجية والمصرية، وبالأرقام الغربية والعربية المشرقية: الأعداد الأصلية والترتيبية والمبالغ والتواريخ الهجرية والميلادية والأوقات وأرقام الهواتف وأرقام الهوية والآيبان والنسب المئوية والمقاييس. ومعها جداول مرجعية ومدققات للصيغ التي تتضمنها الأزواج: أرقام الهوية السعودية والإماراتية والمصرية، وأرقام الهوية الكويتية والبحرينية (البنية فقط)، وبنى الآيبان وخطط ترقيم الهواتف في دول الجامعة العربية، وحروف اللوحات السعودية، وجدول للتحويل من تقويم أم القرى الهجري إلى الميلادي مقارَن بمكتبة ICU، وكتابة الأعداد والمبالغ بالحروف العربية. وكل أرقام الهوية والآيبان والهواتف قيم مولَّدة تجتاز فحوص الصيغة، وليست بيانات شخصية حقيقية.",
      "url": "https://www.bayanatlabs.com/open/gcc-format-validators",
      "url_ar": "https://www.bayanatlabs.com/ar/open/gcc-format-validators",
      "canonical_home": null,
      "licence": "CC-BY-4.0",
      "licence_url": "https://creativecommons.org/licenses/by/4.0/",
      "licence_label": "CC BY 4.0 (test pairs and tables); CC0 1.0 (format facts); Apache-2.0 (phone plans and code)",
      "parts": [
        {
          "name": "ITN/TN test pairs (dev and test splits)",
          "licence": "CC-BY-4.0",
          "licence_url": "https://creativecommons.org/licenses/by/4.0/",
          "in_release": true
        },
        {
          "name": "Compiled format tables, calendar table and number words",
          "licence": "CC-BY-4.0",
          "licence_url": "https://creativecommons.org/licenses/by/4.0/",
          "in_release": true
        },
        {
          "name": "Format facts (lengths, prefixes, field positions, check-digit rules, letter maps)",
          "licence": "CC0-1.0",
          "licence_url": "https://creativecommons.org/publicdomain/zero/1.0/",
          "in_release": true
        },
        {
          "name": "Phone numbering plans (from libphonenumber metadata)",
          "licence": "Apache-2.0",
          "licence_url": "https://www.apache.org/licenses/LICENSE-2.0",
          "in_release": true
        },
        {
          "name": "Validators and scorer (Python package)",
          "licence": "Apache-2.0",
          "licence_url": "https://www.apache.org/licenses/LICENSE-2.0",
          "in_release": true
        }
      ],
      "version": "v0.1.0",
      "date_published": "2026-10-15",
      "date_modified": "2026-10-15",
      "families": [
        "voice",
        "documents"
      ],
      "modality": "structured",
      "languages": [
        "ar",
        "en"
      ],
      "key_numbers": {
        "pairs": 2000,
        "dev-pairs": 600,
        "test-pairs": 1400,
        "classes": 10,
        "iban-structures": 21,
        "phone-plans": 22,
        "hijri-days": 55991,
        "number-words": 8404
      },
      "doi": null,
      "version_doi": null,
      "github": null,
      "huggingface": null,
      "pypi": null,
      "kaggle": null,
      "downloads": [
        {
          "label": "Dataset on Hugging Face",
          "url": "https://huggingface.co/datasets/Bayanat-Labs/gcc-format-validators",
          "kind": "hf"
        },
        {
          "label": "Validators and scorer on GitHub",
          "url": "https://github.com/Bayanat-Labs/gcc-format-validators",
          "kind": "repo"
        }
      ],
      "citation": [
        "Bayanat Labs (2026). GCC Format Validators and Arabic ITN/TN Test Pairs, version 0.1.0. Bayanat Labs. https://huggingface.co/datasets/Bayanat-Labs/gcc-format-validators"
      ],
      "author": "Bayanat Labs"
    },
    {
      "slug": "arabic-document-atlas",
      "type": "dataset",
      "status": "live",
      "title": "Bayanat Labs Arabic Document Atlas",
      "title_ar": "أطلس الوثائق العربية من Bayanat Labs",
      "summary": "Arabic Document Atlas is a sourced training collection for Arabic OCR and document-processing research. The frozen v0.3.0 release contains 200,286 unique decoded images across its image components. The pages configuration packages 199,997 retained page images with Apple Vision machine OCR, approximate word geometry and page-level lineage from 679 selected Arabic Collections Online source book records whose catalog edition dates pass the pre1931 filter. The coverage report distinguishes complete, partial and error processing states; a book with an error can still have published pages. Three exact-pixel duplicate source occurrences are retained separately as lineage, without new images. The reference_pairs configuration supplies 289 source-associated PDF page renders from two PDFs with community-proofread or validated body text, without machine OCR or word boxes; association is not independently verified alignment. The separate community_text configuration contains 26,629 Arabic Wikisource records and source quality labels, without image bytes. Reference transcriptions are already part of that text collection. All four configurations use a training split.",
      "summary_ar": "أطلس الوثائق العربية مجموعة موثّقة المصدر لتدريب أنظمة التعرّف البصري على الحروف العربية ولأبحاث معالجة الوثائق. يضم الإصدار الثابت v0.3.0 عدد 200,286 صورة فريدة بحسب بصمات البكسلات بعد فك الترميز عبر مجموعات الصور. وتضم مجموعة pages عدد 199,997 صورة صفحة محتفظًا بها مع نصوص مستخرجة آليًا بواسطة Apple Vision وإحداثيات تقريبية للكلمات وسجل مصدر لكل صفحة، من 679 سجلًا لكتب مختارة في المجموعات العربية على الإنترنت تستوفي تواريخ طبعاتها المسجّلة في الفهرس شرط النشر قبل عام 1931. ويميّز تقرير التغطية بين حالات المعالجة الكاملة والجزئية والخطأ؛ وقد يحتفظ سجل الكتاب بصفحات منشورة رغم تسجيل خطأ في معالجته. ويُحتفظ بثلاثة مواضع مصدر إضافية لصور مطابقة في البكسلات في مجموعة مستقلة لتتبّع المصدر، دون إضافة صور جديدة. وتوفّر مجموعة reference_pairs عدد 289 صورة صفحة من ملفّي PDF مرتبطة بنصوص المتن ذات وسوم المراجعة أو التحقّق المجتمعي، دون نصوص OCR آلية أو إحداثيات كلمات؛ ولا تمثّل هذه الروابط تحققًا مستقلًا من المطابقة. وتضم مجموعة community_text المستقلة 26,629 سجلًا من ويكي مصدر العربية ووسوم الجودة الأصلية، دون صور. والنصوص المرجعية جزء من هذه المجموعة النصية أصلًا. وتُتاح المجموعات الأربع ضمن قسم train المخصّص للتدريب.",
      "url": "https://www.bayanatlabs.com/open/arabic-document-atlas",
      "url_ar": "https://www.bayanatlabs.com/ar/open/arabic-document-atlas",
      "canonical_home": null,
      "licence": "LicenseRef-Component-Specific",
      "licence_url": "https://huggingface.co/datasets/Bayanat-Labs/arabic-document-atlas/resolve/v0.3.0/DATA_LICENSE.md",
      "licence_label": "Component-specific rights: institution-asserted public-domain ACO images; CC BY 4.0 annotations/docs; CC BY-SA 4.0 community text and reference pairs; Apache-2.0 code",
      "parts": [
        {
          "name": "ACO source page images",
          "licence": "LicenseRef-Component-Specific",
          "licence_url": "https://huggingface.co/datasets/Bayanat-Labs/arabic-document-atlas/resolve/v0.3.0/DATA_LICENSE.md",
          "in_release": true
        },
        {
          "name": "Bayanat Labs annotation representation and documentation",
          "licence": "CC-BY-4.0",
          "licence_url": "https://creativecommons.org/licenses/by/4.0/",
          "in_release": true
        },
        {
          "name": "Arabic Wikisource community text, without images",
          "licence": "CC-BY-SA-4.0",
          "licence_url": "https://creativecommons.org/licenses/by-sa/4.0/",
          "in_release": true
        },
        {
          "name": "Wikisource PDF renders and associated reference body text",
          "licence": "CC-BY-SA-4.0",
          "licence_url": "https://creativecommons.org/licenses/by-sa/4.0/",
          "in_release": true
        },
        {
          "name": "Acquisition, processing and validation pipeline",
          "licence": "Apache-2.0",
          "licence_url": "https://www.apache.org/licenses/LICENSE-2.0",
          "in_release": true
        }
      ],
      "version": "v0.3.0",
      "date_published": "2026-10-15",
      "date_modified": "2026-10-17",
      "families": [
        "documents",
        "vision",
        "knowledge"
      ],
      "modality": "multimodal",
      "languages": [
        "ar"
      ],
      "key_numbers": {
        "unique-images": 200286,
        "machine-word-records": 60699167,
        "aco-page-images": 199997,
        "aco-source-books": 679,
        "reference-pairs": 289,
        "community-text": 26629,
        "mechanical-screen-pass": 188473,
        "aco-catalog-year-cutoff": 1931,
        "complete-books": 389,
        "partial-books": 287,
        "excluded-books": 89,
        "enrolled-books": 768,
        "error-status-books": 92
      },
      "doi": null,
      "version_doi": null,
      "github": "https://github.com/Bayanat-Labs/arabic-document-atlas",
      "huggingface": "https://huggingface.co/datasets/Bayanat-Labs/arabic-document-atlas",
      "pypi": null,
      "kaggle": null,
      "downloads": [
        {
          "label": "Versioned dataset on Hugging Face",
          "url": "https://huggingface.co/datasets/Bayanat-Labs/arabic-document-atlas/tree/v0.3.0",
          "kind": "hf"
        },
        {
          "label": "Explore page images, OCR and source evidence",
          "url": "https://huggingface.co/spaces/Bayanat-Labs/arabic-document-atlas",
          "kind": "hf"
        },
        {
          "label": "Pipeline and release evidence on GitHub",
          "url": "https://github.com/Bayanat-Labs/arabic-document-atlas/releases/tag/v0.3.0",
          "kind": "repo"
        },
        {
          "label": "Component rights and source terms",
          "url": "https://huggingface.co/datasets/Bayanat-Labs/arabic-document-atlas/resolve/v0.3.0/DATA_LICENSE.md",
          "kind": "file"
        },
        {
          "label": "Frozen release statistics",
          "url": "https://huggingface.co/datasets/Bayanat-Labs/arabic-document-atlas/resolve/v0.3.0/metadata/release-summary.json",
          "kind": "file"
        },
        {
          "label": "Shard manifest and checksums",
          "url": "https://huggingface.co/datasets/Bayanat-Labs/arabic-document-atlas/resolve/v0.3.0/metadata/shards.json",
          "kind": "file"
        },
        {
          "label": "Coverage and processing limits",
          "url": "https://huggingface.co/datasets/Bayanat-Labs/arabic-document-atlas/resolve/v0.3.0/reports/coverage.json",
          "kind": "file"
        },
        {
          "label": "Remote file-verification report",
          "url": "https://huggingface.co/datasets/Bayanat-Labs/arabic-document-atlas/resolve/v0.3.0/reports/remote-release-verification.json",
          "kind": "file"
        },
        {
          "label": "Source attribution and credits",
          "url": "https://huggingface.co/datasets/Bayanat-Labs/arabic-document-atlas/resolve/v0.3.0/ATTRIBUTION.md",
          "kind": "file"
        },
        {
          "label": "First pages Parquet shard (partial dataset)",
          "url": "https://huggingface.co/datasets/Bayanat-Labs/arabic-document-atlas/resolve/v0.3.0/data/pages/shard-000000.parquet",
          "kind": "file"
        },
        {
          "label": "First community-text Parquet shard (partial dataset)",
          "url": "https://huggingface.co/datasets/Bayanat-Labs/arabic-document-atlas/resolve/v0.3.0/data/community_text/part-00000.parquet",
          "kind": "file"
        },
        {
          "label": "Reference-pair Parquet from one source PDF (partial configuration)",
          "url": "https://huggingface.co/datasets/Bayanat-Labs/arabic-document-atlas/resolve/v0.3.0/data/reference_pairs/9fce30d8a96a4151b54f0ddd82d2e4d838cb0336.parquet",
          "kind": "file"
        },
        {
          "label": "One duplicate-lineage Parquet shard (partial configuration; no images)",
          "url": "https://huggingface.co/datasets/Bayanat-Labs/arabic-document-atlas/resolve/v0.3.0/data/duplicate_occurrences/shard-000547.parquet",
          "kind": "file"
        }
      ],
      "citation": [
        "Bayanat Labs (2026). Arabic Document Atlas, v0.3.0. Bayanat Labs. https://www.bayanatlabs.com/open/arabic-document-atlas"
      ],
      "author": "Bayanat Labs"
    },
    {
      "slug": "arabic-pii-pack",
      "type": "tool",
      "status": "live",
      "title": "Bayanat Labs Arabic PII Pack for Presidio",
      "title_ar": "حزمة البيانات الشخصية العربية لـ Presidio من Bayanat Labs",
      "summary": "arabic-pii-pack adds Arabic and Gulf personal-data recognisers to Microsoft Presidio and redacts what they find in Arabic and Arabic-English text. Rule-based recognisers find Saudi national ID and Iqama numbers, Emirates ID numbers, Egyptian national numbers, Kuwaiti civil and Bahraini personal numbers, IBANs of the Arab states, phone numbers of the Arab League states, Saudi and UAE vehicle plates, e-mail addresses and URLs, in Western, Arabic-Indic and Persian digits, and check each candidate against its format: check digits, the IBAN registry, birth-date and governorate structure or the phone number plans. An optional model finds Arabic person names. A clitic-aware Arabic context enhancer counts words such as وجوالي and بالهوية as the context words جوال and هوية. What is found can be replaced with a typed placeholder, a mask or a generated value of the same format. The release includes a command line, a Python API that plugs into any Presidio AnalyzerEngine, a synthetic evaluation set and a free in-browser redactor.",
      "summary_ar": "تضيف حزمة arabic-pii-pack إلى Microsoft Presidio أدوات تعرّف على البيانات الشخصية العربية والخليجية، وتخفي ما تجده في النصوص العربية والنصوص المختلطة بالإنجليزية. تكتشف أدوات التعرّف القائمة على القواعد رقم الهوية الوطنية السعودية ورقم الإقامة، ورقم الهوية الإماراتية، والرقم القومي المصري، والرقم المدني الكويتي، والرقم الشخصي البحريني، وأرقام الآيبان للدول العربية، وأرقام الهواتف لدول الجامعة العربية، ولوحات المركبات السعودية والإماراتية، والبريد الإلكتروني، والروابط، بالأرقام الغربية والعربية الهندية والفارسية، وتتحقق من صيغة كل رقم مرشّح: خانة التحقق، أو سجل الآيبان، أو بنية تاريخ الميلاد والمحافظة، أو خطط الترقيم الهاتفية. ويكتشف نموذج اختياري أسماء الأشخاص العربية. ويحتسب معزِّز سياق عربي يراعي اللواصق كلماتٍ مثل «وجوالي» و«بالهوية» على أنها كلمتا السياق «جوال» و«هوية». ويمكن استبدال ما يُكتشف بوسم يذكر نوعه، أو بالإخفاء، أو بقيمة مولَّدة بالصيغة نفسها. ويضم الإصدار أداة لسطر الأوامر، وواجهة Python تعمل مع أي AnalyzerEngine في Presidio، ومجموعة تقييم اصطناعية، وأداة إخفاء مجانية تعمل في المتصفح.",
      "url": "https://www.bayanatlabs.com/open/arabic-pii-pack",
      "url_ar": "https://www.bayanatlabs.com/ar/open/arabic-pii-pack",
      "canonical_home": null,
      "licence": "Apache-2.0",
      "licence_url": "https://www.apache.org/licenses/LICENSE-2.0",
      "licence_label": "Apache-2.0 (code and in-browser redactor); CC BY 4.0 (evaluation set, results and documentation)",
      "parts": [
        {
          "name": "Python package, command line and JavaScript port",
          "licence": "Apache-2.0",
          "licence_url": "https://www.apache.org/licenses/LICENSE-2.0",
          "in_release": true
        },
        {
          "name": "Synthetic evaluation set, results and documentation",
          "licence": "CC-BY-4.0",
          "licence_url": "https://creativecommons.org/licenses/by/4.0/",
          "in_release": true
        },
        {
          "name": "In-browser redactor on Hugging Face Spaces",
          "licence": "Apache-2.0",
          "licence_url": "https://www.apache.org/licenses/LICENSE-2.0",
          "in_release": true
        },
        {
          "name": "Optional name model (CAMeLBERT MSA NER by CAMeL Lab), downloaded at run time",
          "licence": "Apache-2.0",
          "licence_url": "https://www.apache.org/licenses/LICENSE-2.0",
          "in_release": false
        }
      ],
      "version": "v0.1.0",
      "date_published": "2026-10-15",
      "date_modified": "2026-10-15",
      "families": [
        "documents",
        "knowledge"
      ],
      "modality": "tool",
      "languages": [
        "ar",
        "en"
      ],
      "key_numbers": {
        "test-sentences": 1000,
        "rule-spans-found": 1002,
        "rule-spans-missed": 3,
        "rule-false-flags": 1,
        "person-names": 378,
        "person-names-found": 310,
        "wikipedia-paragraphs": 5000,
        "wikipedia-flags": 0
      },
      "doi": null,
      "version_doi": null,
      "github": null,
      "huggingface": null,
      "pypi": null,
      "kaggle": null,
      "downloads": [
        {
          "label": "Open the redactor",
          "url": "https://www.bayanatlabs.com/tools/arabic-pii",
          "kind": "page"
        },
        {
          "label": "Code on GitHub",
          "url": "https://github.com/Bayanat-Labs/arabic-pii-pack",
          "kind": "repo"
        },
        {
          "label": "Evaluation set and results on Hugging Face",
          "url": "https://huggingface.co/datasets/Bayanat-Labs/arabic-pii-eval",
          "kind": "hf"
        },
        {
          "label": "In-browser redactor on Hugging Face Spaces",
          "url": "https://huggingface.co/spaces/Bayanat-Labs/arabic-pii-redactor",
          "kind": "hf"
        }
      ],
      "citation": [
        "Bayanat Labs (2026). arabic-pii-pack: Arabic and Gulf personal-data recognisers and redaction for Microsoft Presidio, version 0.1.0. Bayanat Labs. https://github.com/Bayanat-Labs/arabic-pii-pack"
      ],
      "author": "Bayanat Labs"
    },
    {
      "slug": "arabic-token-tax-index",
      "type": "benchmark",
      "status": "live",
      "title": "Bayanat Labs Arabic Token-Tax Index",
      "title_ar": "مؤشر ضريبة الترميز العربية من Bayanat Labs",
      "summary": "LLM APIs bill and limit by tokens, and the same content usually takes more tokens in Arabic than in English. This index measures how many more, tokeniser by tokeniser, on parallel Arabic–English text from the United Nations: the Arabic/English token ratio with a bootstrap confidence interval, tokens per thousand words, characters per token and a diacritics round trip, and the cost of a page at the vendors' list prices. Short everyday sentences from Tatoeba in Standard, Egyptian and Algerian Arabic are measured too. It comes as a dataset with per-segment token counts, a leaderboard, and a harness that counts the tokens of any text and re-runs the measurement from the public sources.",
      "summary_ar": "تحتسب واجهات النماذج اللغوية الكبيرة تكلفتها وحدودها بالرموز (التوكنات)، وعادةً ما يحتاج المحتوى نفسه إلى رموز أكثر بالعربية منه بالإنجليزية. يقيس هذا المؤشر مقدار ذلك لكل مُرمِّز على حدة، على نص عربي-إنجليزي متوازٍ من الأمم المتحدة: نسبة رموز العربية إلى الإنجليزية مع فترة ثقة بطريقة إعادة المعاينة، وعدد الرموز لكل ألف كلمة، وعدد الحروف لكل رمز، وسلامة التشكيل بعد الترميز وفكّه، وتكلفة الصفحة بأسعار القائمة لدى المزوّدين. ويقيس أيضًا جملًا يومية قصيرة من Tatoeba بالعربية الفصحى والمصرية والجزائرية. ويأتي في صورة مجموعة بيانات بأعداد الرموز لكل مقطع، ولوحة ترتيب، وأداة قياس تحسب رموز أي نص وتعيد القياس كاملًا من المصادر العامة.",
      "url": "https://www.bayanatlabs.com/open/arabic-token-tax-index",
      "url_ar": "https://www.bayanatlabs.com/ar/open/arabic-token-tax-index",
      "canonical_home": null,
      "licence": "CC-BY-4.0",
      "licence_url": "https://creativecommons.org/licenses/by/4.0/",
      "licence_label": "CC BY 4.0 (index data and documentation); Apache-2.0 (harness code and leaderboard); UN sentence pairs under the United Nations Parallel Corpus disclaimer",
      "parts": [
        {
          "name": "Index results, tokeniser metadata, costs, prices and diacritics probe",
          "licence": "CC-BY-4.0",
          "licence_url": "https://creativecommons.org/licenses/by/4.0/",
          "in_release": true
        },
        {
          "name": "Cleaned UN sentence pairs with per-segment token counts",
          "licence": null,
          "licence_url": "https://www.un.org/dgacm/en/content/uncorpus",
          "in_release": true
        },
        {
          "name": "Harness: Python package and command line",
          "licence": "Apache-2.0",
          "licence_url": "https://www.apache.org/licenses/LICENSE-2.0",
          "in_release": true
        },
        {
          "name": "Leaderboard on Hugging Face Spaces",
          "licence": "Apache-2.0",
          "licence_url": "https://www.apache.org/licenses/LICENSE-2.0",
          "in_release": true
        }
      ],
      "version": "v0.1.0",
      "date_published": "2026-10-15",
      "date_modified": "2026-10-15",
      "families": [
        "bench",
        "knowledge"
      ],
      "modality": "benchmark",
      "languages": [
        "ar",
        "en"
      ],
      "key_numbers": {
        "tokenisers": 12,
        "median-ratio": 1.42,
        "lowest-ratio": 0.93,
        "highest-ratio": 3.35,
        "un-test-pairs": 3990,
        "slices": 5,
        "list-prices": 7,
        "vendors": 3,
        "not-measured": 7
      },
      "doi": null,
      "version_doi": null,
      "github": null,
      "huggingface": null,
      "pypi": null,
      "kaggle": null,
      "downloads": [
        {
          "label": "Dataset on Hugging Face",
          "url": "https://huggingface.co/datasets/Bayanat-Labs/arabic-token-tax-index",
          "kind": "hf"
        },
        {
          "label": "Leaderboard on Hugging Face Spaces",
          "url": "https://huggingface.co/spaces/Bayanat-Labs/arabic-token-tax-index",
          "kind": "hf"
        },
        {
          "label": "Harness and index files on GitHub",
          "url": "https://github.com/Bayanat-Labs/arabic-token-tax-index",
          "kind": "repo"
        },
        {
          "label": "Method",
          "url": "https://github.com/Bayanat-Labs/arabic-token-tax-index/blob/v0.1.0/docs/METHOD.md",
          "kind": "repo"
        },
        {
          "label": "Count the tokens of your own Arabic text",
          "url": "https://www.bayanatlabs.com/tools/arabic-token-tax",
          "kind": "page"
        }
      ],
      "citation": [
        "Bayanat Labs (2026). Arabic Token-Tax Index, version 0.1.0 (run 2026-10-16). Bayanat Labs. https://huggingface.co/datasets/Bayanat-Labs/arabic-token-tax-index"
      ],
      "author": "Bayanat Labs"
    },
    {
      "slug": "arabic-asr-channel-pairs",
      "type": "dataset",
      "status": "live",
      "title": "Bayanat Labs Arabic ASR Channel Pairs",
      "title_ar": "أزواج القنوات للتعرّف على الكلام العربي من Bayanat Labs",
      "summary": "How much does a telephone line, a codec or an echoey room hurt Arabic speech recognition? This test set answers it with paired audio: read-speech recordings from FLEURS Egyptian Arabic, each released clean and rendered under fixed, documented channel conditions: a telephone band-limit, down-sampling, a mu-law telephone codec, low-bitrate Opus and synthetic room reverberation. Every variant keeps its parent's reference transcript and is aligned and level-matched to it, so any change in word error rate comes from the channel. The release includes the renderer, the conditions file, synthetic room impulse responses, and word error rates for two open Whisper models scored with ArabicWER. The conditions are simulated on read speech; they are not real telephone or call-centre recordings.",
      "summary_ar": "إلى أي حد يضرّ خط الهاتف أو مرمّز الصوت أو صدى الغرفة بالتعرّف على الكلام العربي؟ تجيب مجموعة الاختبار هذه بتسجيلات صوتية في أزواج: تسجيلات كلام مقروء من العربية المصرية في FLEURS، يُنشر كل منها نظيفًا ومعالَجًا بظروف قنوات ثابتة وموثّقة: حصر نطاق الهاتف، وخفض معدل العيّنات، ومرمّز هاتفي بطريقة mu-law، وترميز Opus بمعدل بتات منخفض، وصدى غرفة محاكى. ويحتفظ كل تسجيل معالَج بالنص المرجعي لتسجيله الأصل، ويُحاذى معه ويُعادَل مستوى صوته، فيرجع أي تغيّر في معدل خطأ الكلمات إلى القناة وحدها. ويضم الإصدار أداة المعالجة، وملف الظروف، واستجابات نبضية اصطناعية لغرف، ومعدلات خطأ الكلمات لنموذجين مفتوحين من Whisper محسوبة بأداة ArabicWER. والظروف محاكاة على كلام مقروء، وليست تسجيلات هاتفية أو تسجيلات مراكز اتصال حقيقية.",
      "url": "https://www.bayanatlabs.com/open/arabic-asr-channel-pairs",
      "url_ar": "https://www.bayanatlabs.com/ar/open/arabic-asr-channel-pairs",
      "canonical_home": null,
      "licence": "CC-BY-4.0",
      "licence_url": "https://creativecommons.org/licenses/by/4.0/",
      "licence_label": "CC BY 4.0 (audio, metadata and measurements); CC BY-SA 4.0 (transcript text); CC0 1.0 (conditions file and impulse responses); Apache-2.0 (renderer code)",
      "parts": [
        {
          "name": "Audio, clean and rendered (derived from FLEURS)",
          "licence": "CC-BY-4.0",
          "licence_url": "https://creativecommons.org/licenses/by/4.0/",
          "in_release": true
        },
        {
          "name": "Metadata, measurements and word error rate tables",
          "licence": "CC-BY-4.0",
          "licence_url": "https://creativecommons.org/licenses/by/4.0/",
          "in_release": true
        },
        {
          "name": "Transcript text: FLoRes-101 reference sentences and the model transcripts",
          "licence": "CC-BY-SA-4.0",
          "licence_url": "https://creativecommons.org/licenses/by-sa/4.0/",
          "in_release": true
        },
        {
          "name": "Conditions file and synthetic room impulse responses",
          "licence": "CC0-1.0",
          "licence_url": "https://creativecommons.org/publicdomain/zero/1.0/",
          "in_release": true
        },
        {
          "name": "Renderer code (arabic-channel-sim)",
          "licence": "Apache-2.0",
          "licence_url": "https://www.apache.org/licenses/LICENSE-2.0",
          "in_release": true
        }
      ],
      "version": "v0.1.0",
      "date_published": "2026-10-15",
      "date_modified": "2026-10-15",
      "families": [
        "voice",
        "bench"
      ],
      "modality": "speech",
      "languages": [
        "ar"
      ],
      "key_numbers": {
        "parents": 500,
        "test-parents": 428,
        "dev-parents": 72,
        "conditions": 5,
        "variants": 2500,
        "clips": 3000,
        "impulse-responses": 60,
        "wer-recordings": 100
      },
      "doi": null,
      "version_doi": null,
      "github": null,
      "huggingface": null,
      "pypi": null,
      "kaggle": null,
      "downloads": [
        {
          "label": "Dataset on Hugging Face",
          "url": "https://huggingface.co/datasets/Bayanat-Labs/arabic-asr-channel-pairs",
          "kind": "hf"
        },
        {
          "label": "Renderer on GitHub (arabic-channel-sim)",
          "url": "https://github.com/Bayanat-Labs/arabic-channel-sim",
          "kind": "repo"
        },
        {
          "label": "Conditions file",
          "url": "https://github.com/Bayanat-Labs/arabic-channel-sim/blob/v0.1.0/src/arabic_channel_sim/data/conditions_v1.yaml",
          "kind": "repo"
        },
        {
          "label": "ArabicWER, the scorer used for the tables",
          "url": "https://www.bayanatlabs.com/tools/arabicwer",
          "kind": "page"
        }
      ],
      "citation": [
        "Bayanat Labs (2026). Arabic ASR Channel Pairs, version 0.1.0. Bayanat Labs. https://huggingface.co/datasets/Bayanat-Labs/arabic-asr-channel-pairs"
      ],
      "author": "Bayanat Labs"
    },
    {
      "slug": "license-check",
      "type": "tool",
      "status": "live",
      "title": "Bayanat Labs Dataset Licence Compatibility Checker",
      "title_ar": "مدقّق توافق تراخيص البيانات من Bayanat Labs",
      "summary": "Can these datasets be merged? Can a commercial model be trained on them? Which licence can the release carry? The checker answers from the licence texts themselves. Give it your sources, each with its licence or terms, and one of 4 intended uses (internal training, commercial model training, redistributing copies or publishing derived annotations), and it returns a verdict for each source and for the mix (allowed, allowed with obligations, unknown or not allowed), the obligations that follow, the licences a release can use, a split plan when sources conflict, and attribution lines. Each step cites the clause it rests on, with a short quotation and the date the text was retrieved. The rules cover 33 licences and terms, from Creative Commons, Open Data Commons and CDLA licences to government open data terms, research-only agreements and model-output terms, and the same rules run in a web app, a Python library, a command-line tool and a GitHub Action. Information, not legal advice.",
      "summary_ar": "هل يمكن دمج هذه البيانات؟ هل يمكن تدريب نموذج تجاري عليها؟ وما الترخيص الذي يحمله الإصدار؟ يجيب المدقّق من نصوص التراخيص نفسها: أدخل مصادرك، ولكل منها ترخيصه أو شروطه، واختر واحدًا من 4 استخدامات مقصودة (التدريب الداخلي، أو تدريب نموذج تجاري، أو إعادة نشر النسخ، أو نشر التعليقات التوضيحية المشتقة)، فيعيد حكمًا لكل مصدر وللمزيج كله (مسموح، أو مسموح بشروط، أو غير معروف، أو غير مسموح)، والالتزامات المترتبة، والتراخيص الممكنة للإصدار، وخطة تقسيم عند تعارض المصادر، وعبارات النسب. وتستند كل خطوة إلى بند تذكره مع اقتباس قصير وتاريخ استرجاع النص. وتغطي القواعد 33 ترخيصًا وشروطًا، من تراخيص المشاع الإبداعي وOpen Data Commons وCDLA إلى شروط البيانات الحكومية المفتوحة واتفاقيات الاستخدام البحثي وشروط مخرجات النماذج، وتعمل القواعد نفسها في تطبيق على الويب ومكتبة Python وأداة سطر أوامر وإجراء في GitHub. معلومات، وليست استشارة قانونية.",
      "url": "https://www.bayanatlabs.com/open/license-check",
      "url_ar": "https://www.bayanatlabs.com/ar/open/license-check",
      "canonical_home": null,
      "licence": "Apache-2.0",
      "licence_url": "https://www.apache.org/licenses/LICENSE-2.0",
      "licence_label": "Apache-2.0 (code and web app); CC BY 4.0 (rules, test cases, schemas and documentation); quotations from licences and terms stay under their owners' terms",
      "parts": [
        {
          "name": "Rules engine, Python library, command line and GitHub Action",
          "licence": "Apache-2.0",
          "licence_url": "https://www.apache.org/licenses/LICENSE-2.0",
          "in_release": true
        },
        {
          "name": "Rules (rules version 2026.10.19), verdict test cases, JSON Schemas and documentation",
          "licence": "CC-BY-4.0",
          "licence_url": "https://creativecommons.org/licenses/by/4.0/",
          "in_release": true
        },
        {
          "name": "Web app on Hugging Face Spaces, in English and Arabic",
          "licence": "Apache-2.0",
          "licence_url": "https://www.apache.org/licenses/LICENSE-2.0",
          "in_release": true
        },
        {
          "name": "Short quotations from licences, terms and FAQs",
          "licence": null,
          "licence_url": null,
          "in_release": true
        },
        {
          "name": "IBM Plex Sans Arabic font in the web app, unmodified (SIL Open Font License 1.1)",
          "licence": null,
          "licence_url": "https://openfontlicense.org/",
          "in_release": true
        }
      ],
      "version": "v0.1.0",
      "date_published": "2026-10-19",
      "date_modified": "2026-10-19",
      "families": [
        "knowledge",
        "align"
      ],
      "modality": "tool",
      "languages": [
        "en",
        "ar"
      ],
      "key_numbers": {
        "licences-and-terms": 33,
        "uses": 4,
        "obligation-types": 15,
        "release-licences": 12
      },
      "doi": null,
      "version_doi": null,
      "github": null,
      "huggingface": null,
      "pypi": null,
      "kaggle": null,
      "downloads": [
        {
          "label": "Open the checker",
          "url": "https://huggingface.co/spaces/Bayanat-Labs/license-check",
          "kind": "hf"
        },
        {
          "label": "Python, command line and GitHub Action",
          "url": "https://www.bayanatlabs.com/tools/license-check",
          "kind": "page"
        },
        {
          "label": "Code, rules and test cases on GitHub",
          "url": "https://github.com/Bayanat-Labs/license-check",
          "kind": "repo"
        },
        {
          "label": "How verdicts are computed",
          "url": "https://github.com/Bayanat-Labs/license-check/blob/v0.1.0/docs/METHOD.md",
          "kind": "repo"
        }
      ],
      "citation": [
        "Bayanat Labs (2026). Dataset Licence Compatibility Checker (license-check), version 0.1.0, rules version 2026.10.19. Bayanat Labs. https://github.com/Bayanat-Labs/license-check"
      ],
      "author": "Bayanat Labs"
    },
    {
      "slug": "arabic-ai-model-tracker",
      "type": "dataset",
      "status": "live",
      "title": "Bayanat Labs Arabic AI Model Tracker",
      "title_ar": "متتبّع نماذج الذكاء الاصطناعي العربية من Bayanat Labs",
      "summary": "A register of Arabic-centric and Arabic-capable AI model releases from 20 May 2023 to 7 October 2026: text LLMs, vision-language and OCR models, speech recognition, text to speech, embeddings, translation and dialect identification. For each release it records the lab, the date, the modality, the sizes the lab states, the Arabic varieties the lab names (quoted, with a link), the benchmark scores the lab reports, the licence as stated, and whether the weights are open, gated, hosted, announced without release or retired. Each row links to its source, with the retrieval time and the SHA-256 of the archived page. The tracker records facts as the labs state them; it does not score, rank or compare models. Version v0.1.0 is a backfill from the labs' own pages, not a complete census. A crawler lists new candidates from Hugging Face, arXiv and the OpenRouter models list for review, and nothing is added automatically.",
      "summary_ar": "سجلّ لإصدارات نماذج الذكاء الاصطناعي المتمحورة حول العربية أو الداعمة لها من 20 مايو 2023 إلى 7 أكتوبر 2026: النماذج اللغوية الكبيرة، ونماذج الرؤية واللغة، والتعرّف الضوئي على النصوص، والتعرّف على الكلام، وتحويل النص إلى كلام، والتضمينات، والترجمة، وتحديد اللهجات. يسجّل لكل إصدار الجهة المطوِّرة والتاريخ ونوع النموذج والأحجام التي تعلنها الجهة واللهجات أو التنوّعات العربية التي تسمّيها (باقتباس ورابط) ونتائج الاختبارات كما تنشرها الجهة والترخيص كما هو مذكور، وهل الأوزان مفتوحة أو مقيّدة بطلب وصول أو متاحة عبر خدمة مستضافة دون أوزان أو معلن عنها دون إتاحة أو متوقفة. ويرتبط كل صف بمصدره مع وقت الاسترجاع وبصمة SHA-256 للصفحة المؤرشفة. يسجّل المتتبّع الحقائق كما تذكرها الجهات، ولا يقيّم النماذج ولا يرتّبها ولا يقارن بينها. والإصدار v0.1.0 جمعٌ أوّلي من صفحات الجهات نفسها وليس حصرًا شاملًا. وتجمع أداة آلية مرشّحين جددًا من Hugging Face وarXiv وقائمة النماذج في OpenRouter لمراجعتهم، ولا يُضاف شيء تلقائيًا.",
      "url": "https://www.bayanatlabs.com/open/arabic-ai-model-tracker",
      "url_ar": "https://www.bayanatlabs.com/ar/open/arabic-ai-model-tracker",
      "canonical_home": null,
      "licence": "CC-BY-4.0",
      "licence_url": "https://creativecommons.org/licenses/by/4.0/",
      "licence_label": "CC BY 4.0 (register, exports, charts and documentation); Apache-2.0 (crawler and build code); quotations stay with their authors",
      "parts": [
        {
          "name": "Register: releases, checkpoints and lab-reported benchmark tables (CSV and Parquet), charts, Atom feed, latest.json and the candidate queue",
          "licence": "CC-BY-4.0",
          "licence_url": "https://creativecommons.org/licenses/by/4.0/",
          "in_release": true
        },
        {
          "name": "Crawler, candidate scorer, checkpoint snapshot and export builder",
          "licence": "Apache-2.0",
          "licence_url": "https://www.apache.org/licenses/LICENSE-2.0",
          "in_release": true
        },
        {
          "name": "Short quotations from model cards, papers and announcements",
          "licence": null,
          "licence_url": null,
          "in_release": true
        }
      ],
      "version": "v0.1.0",
      "date_published": "2026-10-19",
      "date_modified": "2026-10-19",
      "families": [
        "bench",
        "knowledge"
      ],
      "modality": "structured",
      "languages": [
        "en",
        "ar"
      ],
      "key_numbers": {
        "releases": 89,
        "labs": 29,
        "checkpoints": 214,
        "benchmark-rows": 33,
        "releases-with-scores": 20,
        "open-weights": 61,
        "gated-weights": 22,
        "arabic-centric": 27,
        "naming-a-variety": 26,
        "text-llm": 46,
        "speech-asr": 11,
        "candidates": 128
      },
      "doi": null,
      "version_doi": null,
      "github": null,
      "huggingface": null,
      "pypi": null,
      "kaggle": null,
      "downloads": [
        {
          "label": "Dataset on Hugging Face",
          "url": "https://huggingface.co/datasets/Bayanat-Labs/arabic-ai-model-tracker",
          "kind": "hf"
        },
        {
          "label": "Browse and filter the release table",
          "url": "https://www.bayanatlabs.com/tools/arabic-ai-model-tracker",
          "kind": "page"
        },
        {
          "label": "Atom feed of new releases",
          "url": "https://huggingface.co/datasets/Bayanat-Labs/arabic-ai-model-tracker/resolve/main/feed.atom",
          "kind": "hf"
        },
        {
          "label": "Releases table (CSV)",
          "url": "https://huggingface.co/datasets/Bayanat-Labs/arabic-ai-model-tracker/resolve/main/csv/releases.csv",
          "kind": "file"
        },
        {
          "label": "latest.json: headline counts and recent releases",
          "url": "https://huggingface.co/datasets/Bayanat-Labs/arabic-ai-model-tracker/resolve/main/latest.json",
          "kind": "hf"
        },
        {
          "label": "Code and inclusion rules on GitHub",
          "url": "https://github.com/Bayanat-Labs/arabic-ai-model-tracker",
          "kind": "repo"
        },
        {
          "label": "Inclusion rules",
          "url": "https://github.com/Bayanat-Labs/arabic-ai-model-tracker/blob/v0.1.0/RULES-INCLUSION.md",
          "kind": "repo"
        }
      ],
      "citation": [
        "Bayanat Labs (2026). Bayanat Labs Arabic AI Model Tracker, version 0.1.0 (as of 2026-10-19). Bayanat Labs. https://huggingface.co/datasets/Bayanat-Labs/arabic-ai-model-tracker"
      ],
      "author": "Bayanat Labs"
    }
  ]
}