Our previous piece used the IceCube Nobel to set out the wider landscape: multinational facilities, Chinese research reaching the world, and academic translation becoming a standing need rather than a pre-submission favour. This piece returns to ground level and explains how we actually do the work. Our academic clients fall into four broad groups: university lecturers, hospital clinicians, research-institute teams and corporate R&D centres. All of them ask for “translation”, but the work differs considerably.
First: match by discipline, not by language
The biggest difference between a research paper and an ordinary business document is that a non-specialist genuinely cannot follow it. A condensed-matter paper and a clinical study require almost no shared background knowledge. We assign work by discipline first: translators with research backgrounds in science and engineering handle experimental and technical manuscripts, while translators specialising in medical translation handle clinical records, pharmaceutical materials and study documentation. A translator has to follow the logic of the argument before it can be written accurately in another language.
Second: terminology first, with equations and figures checked item by item
Before translation begins we build a glossary: core concepts, proper nouns and the treatment of abbreviations are agreed with the client. Equation numbers, figure captions, axis labels and references are all in scope. Units are a perennial trouble spot — metric and imperial mixed together, inconsistent significant figures — and both can cause misunderstandings in laboratory documents. Numbers and parameters are treated as zero-tolerance items, with reviewers checking them against the source one by one.
Third: two-stage review, for language and for subject matter
Academic manuscripts pass through two reviews. The language review examines sentence structure, tense, cohesion and academic register — methods are conventionally written in the past tense, established findings in the present — and reorganises sentences rather than carrying them across word by word. The specialist review checks terminology and usage against the conventions of the field, and confirms that conclusions have not shifted during rewriting. Only after both reviews is the manuscript finalised. Reference formatting, layout and figure labelling follow the target journal’s requirements, sparing clients reformatting at submission.
Fourth: confidentiality is a hard rule
Much of what reaches us is unpublished: results not yet in print, proposals not yet disclosed and technical descriptions ahead of patent filings. This material is covered by confidentiality arrangements throughout, restricted to the assigned translators and kept out of any external-tool pipeline, with copies handled as agreed once the project ends. For research clients this matters more than price.
What machines do, and do not, handle
Machine translation has a legitimate place: during literature review, screening large volumes of references, and for repetitive experimental records, an MTPE workflow with a machine draft and human revision is markedly more efficient. But final manuscripts for submission and grant proposals for reporting purposes should not rely on raw machine output; subtle differences in argument and the conventions of a discipline are exactly what machines most easily flatten.
A common companion request
Academic cooperation regularly brings document work in its wake: transcripts, degree certificates, employment letters and invitation letters for visiting scholars need translations for visa applications or university registration. These go through certified translation with an official stamp: page-by-page correspondence with the original and a translation stamp that is well recognised by universities and visa reviewers.
If you have a manuscript, send your field, its purpose (submission, application or reporting) and your deadline through the contact page and we will propose an approach. For the wider background see “IceCube Wins the Nobel Prize: the Language Chain Behind a 12-nation Collaboration”, and “Want to Translate Science? A Skill Checklist Drawn from the IceCube Story” takes the translator’s view.