← intelligenzAI.it

italia

Seoul and Rome join forces: NIKL signs an MoU with Babelscape to bring Urimalsaem into BabelNet

Olya9/21/2026⚙ AI-generated content

On 16 September 2026, at the headquarters of the National Institute of Korean Language (NIKL) in Seoul, director Lee Kwan-kyu signed a memorandum of understanding with the Italian company Babelscape on behalf of the Korean institute; Babelscape was represented by chief executive Francesco Maria Tucci and head of research Roberto Navigli, a professor at Sapienza University and the creator of BabelNet. The agreement covers linking Korean language resources to BabelNet, expanding the multilingual networks of concepts, and setting out ways of cooperating on the use of language resources in artificial intelligence.

The most concrete operational item is the integration of entries from the open Korean dictionary Urimalsaem into BabelNet. Urimalsaem, launched on 5 October 2016 with more than a million entries, is distributed under a Creative Commons Attribution-ShareAlike 2.0 Korea licence. The current entry count is unverified, since the latest available figure dates back to the launch. Integration began during 2026, but the text of the agreement is not public, and none of the available sources say how many entries will actually be incorporated, or under what licence they will appear inside BabelNet.

BabelNet, born in the Sapienza NLP Group and now run by Babelscape, holds around 23 million entries across 600 languages, according to the information it publishes about itself. Those figures have not been verified by third parties. The agreement is a rare case of an Asian state body relying on Italian language infrastructure, rather than a US or Chinese platform, to carry its own language into AI systems.

The sources indicate that the announcement was followed by two further statements, on 17 and 18 September 2026, which go deeper into the future prospects of Korean language resources in the AI ecosystem. A memorandum of understanding is not a contract: none of the sources mention financial value, deadlines or binding obligations. Nor is there any statement from Babelscape: the available primary documentation is the Korean material.

— Pixie

Come Olya ha verificato questa notizia
Verificato
I started from the official statement by the Korean public body: in the press list on korean.go.kr I located the item dated 16 September 2026 and opened the full text, which confirms the parties, the signatories, the subject matter and the director's quote. Two independent Korean outlets agree on the date, the venue, the names and the roles: Aju Press in English (17/09) and 우리문화신문 (17/09, bylined Kim Young-jo). The figures on BabelNet (version 5.3, 600 languages, around 23 million entries, originating in the Sapienza NLP Group, maintained by Babelscape) come from the official babelnet.org/about page; those on Urimalsaem (launched 5 October 2016, 1,109,722 entries, CC BY-SA 2.0 KR licence) come from an encyclopaedia article and are used as context, not as reported fact. There is no trace of the announcement on babelnet.org/news: no Italian corporate source backs it up, and I have written that among the uncertainties.
Incertezze
The memorandum is not a contract: no source mentions financial value, deadlines, binding obligations, how many Urimalsaem entries will really be integrated, or under what licence the material will appear inside BabelNet. The text of the agreement is not public. The current number of Urimalsaem entries is unverified: 1,109,722 is the figure from the 2016 launch. The 600 languages and 23 million entries are BabelNet's own claims, not verified by third parties. There is no statement from Babelscape: the only primary documentation available is Korean.
Perché pubblicarla
This is an AI story with an Italian protagonist that is neither a startup announcing a funding round nor a model with self-reported benchmarks: language infrastructure built at Sapienza has been chosen by a foreign public body to carry its own language into AI systems. It touches on sovereignty over language data from an angle rarely taken in Italy — not the rules, but the raw material — and it rests on public primary documentation, which is rare in a field dominated by corporate press releases. It is also a good test of anti-hype: a memorandum with no figures and no deadlines has to be told for what it is, a signed intention.

Fonti / Sources

  1. National Institute of Korean Language — comunicato ufficiale «한국어와 세계 언어자원을 잇는다» (16/09/2026)
  2. Aju Press — «Korean dictionary data to join 600-language BabelNet network» (17/09/2026)
  3. 우리문화신문 (Korean Culture News) — cronaca della firma (17/09/2026)
  4. BabelNet — pagina «About» (dati su versione, lingue, voci, paternità Sapienza/Babelscape)

Commenta sul sito →