LM- vs Corpus-driven translation of a Special Power of Attorney

Legal language is characterised by peculiar features and fixed expressions such as formulae. The literature has long highlighted the way to cope with the challenges of non-equivalence in legal translation. For example, paraphrasing and functional translation have been considered as acceptable strategies. This paper is aimed at exploring whether and to what extent the formulaic expressions of a special power of attorney written in American English can be rendered into Italian by consulting an LM interface and an ad hoc corpus of parallel texts. The paper findings highlight that full equivalences are difficult to obtain in such formulaic contexts. In addition, the analysis indicates that functional translation and paraphrasing (often suggested by the LM tool) are debatable, especially if the target language features no expressions linguistically and/or pragmatically similar to source phrases. This may be due to gaps between the source and target legal systems that translators are not necessarily called on to bridge. With regard to the reliability of the tools under scrutiny, the LM platform often provides satisfactory language solutions, thereby underscoring that, if integrated with corpus-driven post-editing, it can yield accurate results.