Data Marketplace

Flitto Data Marketplace offers license-verified AI training and evaluation datasets in text, speech, image, and video across 100+ languages and 23+ domains. Try a free sample, then contact us to purchase.

Sell your data on Flitto

Simply register your data and check its sale eligibility.

  • NEWEvaluation DataMultimodal

    Early Devanagari Print Benchmark & Ground-Truth Corpus

    This human-verified ground-truth dataset and evaluation benchmark is built from rare, early 20th-century Devanagari (Hindi) printed books. It provides dual-adjudicated transcriptions and error-scored pairs specifically designed to resolve OCR and VLM discrepancies on challenging historical texts.

    DomainHumanities and Social
    LanguageHindi
  • NEWPre-training DataMultimodal

    Licensed Pashto Print Corpus

    A digital text corpus built from Pashto printed books licensed by publishers and authors. It includes rare lithographed editions and provides original page scans, OCR-generated text, and transcriptions cross-verified by native speakers.

    DomainHumanities and Social
    LanguagePashto
  • NEWEvaluation DataImage

    Historical Urdu Nastaliq OCR & VLM Benchmark

    This held-out benchmark is designed to evaluate OCR and vision-language models on dense, 1908 lithographed Urdu dictionary scans. It features 161 line images paired with human-adjudicated ground truth text, alongside detailed word-level error scoring across frontier AI models.

    DomainHumanities and Social
    LanguageUrdu
  • NEWPre-training DataText

    Industry, Tech, and Global Business In-Depth Articles Dataset

    A Korean-language corpus of in-depth feature articles centered on core business fields such as AI, advanced industry, personal finance, and macroeconomics.

    DomainManagement, Economic and Finance|IT and Tech
    LanguageKorean
  • NEWPre-training DataText

    Newsletter & Email Marketing Campaign Dataset

    This dataset consists of email marketing campaigns sent from multiple newsletters, along with a section-level breakdown of the same emails.

    DomainAd and Marketing
    LanguageEnglish
  • NEWPre-training DataText

    Long-Form Online Media Articles Dataset

    This dataset contains long-form feature articles collected from multiple U.S. online media outlets, spanning a wide range of domains and writing styles.

    DomainLifestyle
    LanguageEnglish
  • NEWPre-training DataText

    Personal Finance Community Forum Dataset

    This dataset consists of real discussions and posts from a personal finance community, covering topics such as budgeting, saving, and debt.

    DomainManagement, Economic and Finance
    LanguageEnglish
  • NEWEvaluation DataText

    Nigeria Centric African Cultural Bias Exploration Dataset

    Adversarial prompts and evaluation datasets designed to detect, assess, and mitigate cultural bias and safety risks in LLMs within African regional contexts.

    DomainHumanities and Social
    LanguageEnglish|Yoruba|Hausa
  • NEWFine-tuning DataText

    African Languages Medical & Clinical Multilingual Parallel Corpus

    Multilingual parallel corpus aligning medical and clinical sentences across English, Hausa, Swahili, Yoruba, Igbo, and Amharic, covering patient instructions and clinical questions used in healthcare settings.

    DomainMedical
    LanguageEnglish|Swahili|Yoruba|Amharic|Hausa
  • NEWFine-tuning DataText

    African Multilingual Customer Service Conversation Dataset

    Multi-turn conversational dataset across Hausa, Yoruba, and Swahili designed for instruction-tuning customer service LLMs.

    DomainLifestyle|Management, Economic and Finance|IT and Tech
    LanguageHausa|English