Evaluating the performance of DeepL as translation tool between German and Italian Easy Language administrative texts

Easy Language (EL), a comprehensibility-optimised form of a natural language that makes content accessible to people with communication impairments, can play an important role in institutional communication. Howev-er, the extent to which it is adopted for legal and administrative texts has not progressed evenly in the various countries. Interlingual translation could be an asset to increase the use of EL in this domain, also considering that recently AI-driven tools have become an important support for translation. The project that will be presented aims to evaluate the possibility of using DeepL as an automated translation tool for interlingual translation into EL particularly in the domain of administrative texts. To this end, the performance of DeepL was analysed for the language pair German and Italian, using a EL bilingual corpus produced by the administration of the Province of Bolzano/Bolzen. The machine-generated corpus texts were evaluated for readability, correctness and compliance with the EL rules and compared with the gold standard texts, in order to identify the potentials and limitations of the translation and the possible need for post-editing.