
thumb|Eurodicautom Eurodicautom was the pioneering terminology database of the European Commission, launched in 1975 to support translators and staff in managing multilingual terminology. Originating from computational linguistics research at the Université libre de Bruxelles, it became one of the first large-scale digital terminology systems, initially covering six languages and expanding to eleven (plus Latin for scientific names) as the European Community grew. By 1980, it was accessible online within the Commission, and later offered public access, influencing modern translation technologi
thumb|Eurodicautom Eurodicautom was the pioneering terminology database of the European Commission, launched in 1975 to support translators and staff in managing multilingual terminology. Originating from computational linguistics research at the Université libre de Bruxelles, it became one of the first large-scale digital terminology systems, initially covering six languages and expanding to eleven (plus Latin for scientific names) as the European Community grew. By 1980, it was accessible online within the Commission, and later offered public access, influencing modern translation technologies such as IATE.
== History and development == === Origins with DICAUTOM === Eurodicautom originated from DICAUTOM (Dictionnaire Automatique), an automated dictionary consultation project developed between 1961 and 1963 by the Groupe de Linguistique Automatique at the Université libre de Bruxelles (ULB) under Euratom contract No. 018615 CETB. This collaborative effort involved researchers such as J.A. Bachrach, J. Blois, P. Decresy, F. Defijn, L. Hirschberg, and J. Mommens. DICAUTOM aimed to assist human translators by automating dictionary searches, utilizing a morphological analysis system developed by Jacques Blois and lexical choice concordances from high-energy physics by L. Hirschberg, implemented on early computers like the IBM 7090 and IBM 1620. It drew inspiration from machine translation research at Georgetown University.
Discovered by embedding cosine similarity (sentence-transformers MiniLM, 384-dim).