Is machine translation reliable in the legal field? A corpus-based critical comparative analysis for teaching ESP at tertiary level

This paper is aimed at exploring whether and how machine translation (MT) can be relied on in the legal field. To this aim, clauses excerpted from a distribution agreement are translated from Italian into English by using the Deepl machine translation platform. In order to assess the reliability and quality of the translation, a corpus of distribution agreements written in English as a first language and as a lingua franca is composed. The contracts are sourced from the Onecle.com platform, and the reference corpus is built semi-automatically by using the BootCaT software solution. In order to retrieve documents drafted in international English and hence validated by international lawyers or businesspeople, advanced search techniques are applied. The corpus is then consulted by using the AntConc offline concordancer, and the MT is compared with corpus evidence. The paper findings highlight shortcomings in the MT related to legal formulae. Word order is sometimes incorrect and the system specificity of the legal language in the target text remains unaddressed. The paper calls for future improvements in MT software, and reports the need for translators and translation students to be acquainted with legal language style and writing conventions