Scientific translation turns complex research and technical evidence into accurate content for another language and audience. The work may involve journal articles, research reports, protocols, laboratory documentation, conference materials, grant applications or public-facing science communication. In every case, terminology, context and traceability matter.

A translator does not need to hold every qualification represented in a multidisciplinary project. They do need the right subject knowledge, strong research skills, access to reliable references and a workflow that brings in scientific or medical reviewers when the risk demands it.

What makes scientific translation different?

Scientific texts compress a great deal of meaning into specialised terminology, symbols, abbreviations and conventional structures. A word may have a precise definition within one discipline and a different meaning elsewhere. Small changes can alter a method, result or conclusion.

The translator must understand the document’s purpose as well as its language. A manuscript submitted to a journal, a protocol used by researchers and an article written for the public require different levels of detail, tone and explanation.

Scientific translation therefore combines:

  • subject-specific language and concepts;
  • target-language scientific conventions;
  • consistent terminology and nomenclature;
  • accurate handling of data, units and references;
  • clear decisions about ambiguity;
  • documented review and quality control.

Match expertise to the scientific field

“Science” is not one subject. Life sciences, chemistry, physics, environmental research, engineering and social science each use different methods and language. Even within medicine, an oncology paper and a medical-device usability study call for different knowledge.

Project scoping should identify the discipline, document type, audience and intended use. That information helps select a linguist with relevant experience and decide whether the translation also needs review by a subject-matter expert.

Credentials alone are not enough. A strong scientific translator knows how to investigate unfamiliar concepts, distinguish authoritative sources from unreliable ones and flag uncertainty instead of guessing.

Terminology needs evidence and control

A general dictionary rarely resolves specialised terminology. Translators may need peer-reviewed literature, institutional databases, standards, product documentation, existing bilingual publications and client-approved glossaries.

Terminology work should record the preferred term, definition, source, context and any prohibited alternatives. This is particularly important when several translators work on a programme or when content is updated over time.

Names of organisms, chemicals, genes, proteins and units may follow formal systems. Capitalisation, italics, symbols and spacing can carry meaning. The target text must follow the conventions used by the relevant discipline and market.

Preserve data, methods and references

Numbers deserve the same attention as words. Decimal separators, measurement units, ranges, confidence intervals, table labels and figure references must remain accurate and consistent. Conversion should only take place when it is required and authorised.

The translator should also verify cross-references, citations, footnotes, equations and labels after layout or reintegration. A correct sentence can still mislead if it points readers to the wrong figure or if a minus sign disappears during formatting.

Source inconsistencies should be queried and documented. Translation is not permission to silently repair a scientific claim, change a result or complete missing information.

Scientific translation and plain-language communication

Research content is often adapted for patients, policy makers, investors, journalists or the general public. This is a different task from translating a specialist paper.

Plain-language science communication can explain technical concepts, define abbreviations and simplify sentence structure, but it must preserve the evidence and its limitations. Uncertainty should not become certainty, and correlation should not become causation.

Agree the audience and reading level before production. If the brief includes both translation and adaptation, keep the two decisions visible so reviewers can understand how the final wording was reached.

Can AI be used for scientific translation?

AI and machine translation can support controlled scientific workflows, especially for repetitive content, terminology extraction and first-pass drafting. They should not be treated as independent scientific authorities.

A safe workflow starts with approved source content, terminology, reference material and clear rules about what the system may change. Outputs then need review by a qualified linguist, with subject-matter escalation for ambiguous, novel or high-risk content.

Public AI tools may also create confidentiality and data-protection risks. Before uploading unpublished research, personal data or proprietary results, check the agreed platform, data processing terms and client policy.

The decision should reflect the document’s use. Internal discovery material and a publication-ready paper do not carry the same risk.

A robust scientific translation workflow

A practical process normally includes:

  1. Scoping: identify the field, audience, purpose, files, languages and risk.
  2. Reference review: collect glossaries, previous publications, style guides and authoritative sources.
  3. Terminology preparation: resolve key concepts and record approved equivalents.
  4. Translation: work in a controlled environment with queries and version tracking.
  5. Revision: review accuracy, clarity, terminology and target-language conventions.
  6. Subject review: involve a domain specialist where scientific judgement is required.
  7. In-context QA: check tables, figures, references, units, layout and final files.

For recurring programmes, translation memories and terminology bases improve consistency, but they still need governance. An old approved sentence may no longer be correct when the underlying research or regulation changes.

Quality checks for scientific content

Final QA should cover more than spelling. Reviewers should verify:

  • meaning, omissions and additions;
  • specialised terminology and abbreviations;
  • numbers, units, symbols and nomenclature;
  • tables, figures and cross-references;
  • citations and bibliography formatting;
  • consistent UK English or the required target variety;
  • tracked queries and authorised decisions;
  • confidentiality and file integrity.

Where the content affects health, safety, regulation or formal publication, define the required validation and sign-off at the start. Quality cannot be added reliably after an unclear workflow.

Choosing a scientific translation partner

Ask how the provider selects linguists, manages terminology, handles confidential material and escalates subject questions. The right team should be able to explain its controls and limits clearly.

Ontranslation builds scientific translation workflows around the real document, audience and risk. We combine specialist linguists, controlled terminology, human revision and in-context QA, with governed AI where it is appropriate.

Tell us about the scientific field, content type, languages and intended use, and we will define the expertise and review process the project requires.

Scientific translation quality checklist

This guide focuses on the editorial and technical checks that make scientific content reliable. Start by identifying the audience and publication context: a research paper, patient-facing summary and technical report require different terminology and explanations.

  • Terminology: use approved nomenclature and document ambiguous terms.
  • Data: verify figures, units, decimal conventions and statistical notation.
  • Evidence: preserve citations, references, captions and cross-references.
  • Visuals: check labels in tables, charts and diagrams.
  • Review: combine linguistic revision with subject-matter validation when risk demands it.

Use AI for support, not scientific validation

AI can help detect inconsistencies or prepare terminology candidates, but fluent output does not prove scientific accuracy. Protect unpublished material, compare every critical statement with the source and assign final responsibility. For a project-specific assessment, share the field, document type, languages and publication requirements.