Errors in post-editing machine-translated legal texts - examples
Recent statistics indicate an increase in the use of machine translation post-editing in the language industry. But despite the rapid advancements in the field of neural machine translation (NMT), subtle errors can still creep in. And these can be particularly damaging in the field of legal translation, where accuracy is everything. In this article, we provide examples of some of the main errors that post-editors working on German-English translations encounter in their work.
Mistranslation of legal terms
Legal terms may not have direct equivalents in the target language. NMT tools can produce translations that do not accurately convey the intended legal meaning. Experienced post-editors might inadvertently accept these translations if the overall document appears coherent.
- The German legal term ‘Eigenbedarf’ (personal use of property) could be wrongly translated as ‘own needs’, which would not convey the correct legal meaning.
False friends are words that look similar in two languages but have different meanings. In legal translation, false friends can lead to serious misunderstandings. Experienced post-editors might occasionally miss them, assuming that similar-looking terms mean the same thing in both languages.
- In German, ‘aktuell’ means ‘current’ or ‘up to date’, but the NMT tool might mistranslate it as ‘actual’, which would be confusing to an English reader.
Subtle contextual errors
NMT systems may have difficulty understanding subtle contextual nuances in legal language. Post-editors might miss these errors, especially if the overall meaning of the sentence seems reasonable but the specific legal context has not been accurately captured.
- The German phrase ‘nach billigem Ermessen’ might be translated by an NMT system as ‘at the reasonable discretion’ [of the court]. However, a post-editor might miss the subtle legal nuance and not realise that simply ‘at the discretion’ [of the court] would be a more accurate translation.
Legal language often contains idiomatic expressions and phrases that may not have direct equivalents in the target language. NMT tools may produce literal translations that sound awkward or wrong. Post-editors might miss these nuances, especially if the literal translation appears grammatically correct.
- The German expression ‘unter einer Decke stecken’ (to be in cahoots) could be translated literally by an NMT system, resulting in a non-idiomatic phrase in English (in this case, ‘stuck under a blanket’).
Consistency in terminology across documents
When working on multiple legal documents, NMT tools might introduce inconsistencies in terminology across documents. This can be difficult to spot for post-editors, especially when working on large translation projects.
- The NMT system may inconsistently translate ‘Klage’ as ‘lawsuit’ in one document and as ‘claim’ in another, or ‘Ehevertrag’ as both ‘marriage contract’ and ‘matrimonial agreement’ in the same document, leading to inconsistencies in the terminology of the entire translation project.
Missing legal references
Legal texts often refer to specific laws, regulations or case law. NMT tools don’t always translate these references accurately. Post-editors might miss such omissions, assuming that the references have been created and are accurate.
- The German provision ‘§ 826 BGB’ (Art. 826 Swiss Civil Code) may be omitted or mistranslated by the NMT tool, resulting in the absence of correct legal references in the English translation.
Lack of proper citation
Similarly, legal documents often require precise citation of legal authorities. NMT systems can struggle with citation formats. Post-editors might miss instances of missing or incorrect citations.
- The NMT system may not include proper legal citations or not in the correct format, e.g. failing to cite the specific article or section of a law referenced in the original German text.
Incomplete legal analysis
Some legal documents require a thorough analysis of the law and a full understanding of the legal context and background. As things stand, this is still beyond the capabilities of most NMT tools, which means post-editors have to be extra vigilant when reviewing and editing the machine-generated output.
- The NMT system may produce a translation that lacks the in-depth legal analysis required for a contractual clause, omitting essential details or considerations.
Orthographic and formatting errors
NMT systems can introduce spelling, punctuation and formatting errors. Post-editors might miss minor punctuation or formatting issues because they are more focused on content and meaning.
- The NMT system may introduce punctuation errors, such as missing or misplaced commas, which may affect the legal interpretation of a sentence.
These examples show how machine translation errors in legal documents can manifest in various ways, underscoring the crucial role of professional post-editors in ensuring linguistic precision and conformity to legal norms. Although neural machine translation offers significant benefits and is rapidly improving, it still has its limitations. This means that training, feedback and collaboration between legal and language professionals are essential for refining the quality of machine-translated legal texts over time.
Machine translation post-editing (MTPE) has established itself in the language industry over the past few years and is unlikely to change any time soon. Are you looking for a reliable partner to ensure that your machine translated legal texts are edited according to your needs and specifications? Then don’t hesitate to contact us and we’ll be happy to assist you.