Korean corpora

• Modern Korean

  • Modu Corpus (Everyone's Corpus, NIKL)
    Corpus portal of the National Institute of Korean Language: c. 157 resources — written, newspaper, spoken, regional-dialect (2021), parallel and annotated (morphology, syntax, semantics) corpora. Download after application and signed agreement.
  • Modu Corpus: concordance search (용례 검색기)
    Online example (concordance) search over NIKL corpora; the same portal offers an n-gram analyser and an Open API.
  • NIKL Korean Learner Corpus
    Separate NIKL site for the corpus of learners of Korean (listed on the Modu portal as an externally hosted resource).
  • AI Hub
    AI training-data platform of NIA (Korean Ministry of Science and ICT); its Korean-language section has text and speech datasets, incl. dialect speech (Gyeongsang, Jeolla, Chungcheong, Gangwon, Jeju). Download after registration.
  • Sketch Engine: Korean corpora
    List of Korean corpora in Sketch Engine: web corpora koTenTen12 (461M words) and koTenTen18 (1.67B), JSI Timestamped, CHILDES, OPUS, OpenSubtitles. Paid subscription (some corpora on trial).
  • koTenTen Korean web corpus (Sketch Engine)
    Korean TenTen web corpus in Sketch Engine (2018 version: 1.67B words); subscription or trial access.
  • Universal Dependencies: Korean treebanks
    Five Korean UD treebanks: GSD (80K tokens), Kaist (350K), KSL (learner Korean, 155K), LittlePrince, PUD; open licences.
  • UD Korean-Kaist treebank
    Universal Dependencies treebank converted from the KAIST Tree-Tagging Corpus: 27,363 sentences, 350,090 tokens; CC BY-SA 4.0.
  • Korean parallel corpora (Jungyeul Park)
    Korean–English (news, JHE), Korean–French and North Korean–English parallel corpora (Park, Hong, Cha 2016); CC BY-SA 3.0.

• Historical corpora

  • NIKL Corpus of Korean Historical Materials (국어 역사 자료 말뭉치)
    Part of the Modu Corpus: the 2023 release has 1,153 Hangul documents from the 15th to the early 20th c.; the 2024–2025 releases add 17th–19th-c. letters (언간), early-20th-c. novels, the Jeguk sinmun newspaper etc. Access on application.
  • Open Korean Historical Corpus (OKHC)
    KAIST-led corpus (LREC 2026): 17.7M documents, 5.1B tokens, 7th c.–2025, 19 sources; Classical Chinese, Idu, Middle and Early Modern Korean, North Korean etc. Data on Hugging Face, CC BY-NC 4.0.
  • National Hangeul Museum Archive
    Archive of the National Hangeul Museum: full texts of old Hangul documents, letters, eonhae translations and oral-history transcripts; keyword search.

• Jejuan

  • Jejueo datasets JIT and JSS (Kakao Brain)
    JIT: Jejueo–Korean parallel corpus of 170K+ sentences from interview transcripts; JSS: 10K audio files by one native speaker with transcripts. Download via Kaggle.
  • A multi-modal documentation of Jejuan conversations (ELAR)
    Soung-U Kim's collection in the Endangered Languages Archive (ELAR, SOAS): annotated audio-video corpus of spoken Jejuan, mainly conversations, plus songs, rituals, narratives. Some files require login.