This is the presentation delivered at the MDTT 2025 conference in Thessaloniki, showcasing the work carried out within the ATRIUM project on the translation of textual keywords from scientific publications. We proposed a methodology and a functional implementation that leverages Large Language Models to map keywords to entities in multilingual knowledge bases and controlled vocabularies, particularly Wikidata. By integrating these sources, the approach not only enables multilingual keyword translation but also maps terms to Linked Data entities, thereby disambiguating their meaning and enhancing the identification and classification of the related publications.