01 · The Question
Do you need another version of your search strategy for every relevant language?
You know that relevant research may be published in languages other than English. The obvious response seems to be translating your keywords into those languages and adding them to the search.
Sometimes that is exactly what improves retrieval. In other situations, it adds little because the database already exposes English translations, English abstracts, or standardized subject headings. Worse, a literal translation can create false confidence if it does not match the terminology researchers actually use.
The decision is therefore not simply whether translation is good practice. It is whether translated terminology gives your search access to relevant records that your existing strategy would otherwise struggle to retrieve.
03 · What You Need to Know
A multilingual search is a terminology problem, not a translation exercise
First ask whether the database already gives you an English route to the record
Before translating anything, determine what fields your database searches. A non-English publication may already have an English translated title, English abstract, author keywords, or standardized indexing terms.
PubMed, for example, displays English titles and abstracts when they are supplied and separately provides access to original or transliterated non-English titles. Its Title/Abstract search includes titles, abstracts, other abstracts, and author keywords.
If your relevant non-English records are consistently represented through searchable English metadata, translating every term may add little. If important records expose only original-language terminology, the argument for translation becomes much stronger.
Do not translate words before identifying concepts
A search strategy should represent concepts, not merely a list of English words. Suppose your concept is secondary education. The English strategy may already include terms such as secondary school, high school, secondary education, and relevant controlled vocabulary. Translating only one of those phrases into another language does not reproduce the concept.
A better approach is to define the concept first and then identify how that concept is expressed in the target research literature.
Concept Define precisely what the search needs to represent.
English terminology Identify synonyms, spelling variants, abbreviations, and controlled vocabulary already used in the main strategy.
Target-language terminology Identify equivalent and locally used expressions rather than translating one preferred English phrase mechanically.
Validation Test whether the resulting terms retrieve known relevant records in the target source.
Literal translation may not reflect scholarly usage
Languages do not map onto one another word for word. A technically correct translation may be uncommon in scholarly writing, while researchers in the target setting use another expression entirely.
This becomes particularly important for educational stages, health conditions, professions, policy concepts, institutional structures, culturally specific phenomena, and constructs whose definitions vary across research traditions.
Even closely related languages may use different conventional terminology. Within one language, researchers in different countries may also prefer different terms. Translation therefore requires attention to disciplinary and geographic usage, not just lexical equivalence.
Known relevant papers are excellent terminology sources
If you already have relevant papers in the target language, treat them as evidence about how researchers describe the concept. Examine their titles, abstracts, keywords, subject headings, and references.
Search development is inherently iterative. Cochrane's guidance describes search strategies as being modified according to what retrieval reveals and recommends combining text words with controlled vocabulary in databases that provide both.
A useful translated term should therefore prove itself in retrieval. If it cannot find a known relevant article in a field where you expect that term to appear, investigate why.
Controlled vocabulary can reduce some translation problems, but not all
Databases that assign standardized subject headings provide another route into the literature. A subject heading represents a concept consistently even when authors use different words in titles and abstracts.
This can reduce dependence on translated free-text terminology. It does not eliminate it. Not every record is fully indexed, new terminology may not yet map cleanly onto established headings, and databases differ in their thesauri and indexing practices. Cochrane consequently recommends combining controlled vocabulary and text-word searching where appropriate.
Translation is especially useful in regional and local discovery systems
The case for translated search terms becomes stronger when you move beyond international databases into sources designed primarily for a particular country, region, or language.
Journal websites, repositories, library catalogues, government portals, and regional databases may expose original-language metadata more prominently than English metadata. If your question is geographically focused, searching appropriate regional databases and adapting the terminology to those systems may reveal research absent from your initial search.
Original scripts and transliteration require attention
Languages written in non-Roman scripts create another layer of search design. A database may store an original-script title, a romanized or transliterated form, an English translation, or some combination of these.
PubMed illustrates why field behavior matters. The National Library of Medicine notes that original non-English titles can be searched through its Transliterated Title field, while its default title searching uses English titles and title translations.
Do not assume that entering original-script terminology, romanization, and English translation into the same search box produces equivalent retrieval. Check the documentation of the actual platform you are using.
Translation quality matters more when a concept is ambiguous
A simple concrete term may be relatively straightforward to translate. An abstract construct such as belonging, resilience, engagement, social capital, inclusion, or academic persistence can be much harder because its meaning depends on theoretical and disciplinary context.
For consequential searches, consult someone who understands both the language and the research domain where feasible. A fluent speaker who does not know the discipline may choose everyday vocabulary that researchers rarely use; a subject specialist with weak language proficiency may miss important semantic distinctions.
Machine translation can generate candidates, but candidates still need testing
Translation software can be useful for producing preliminary equivalents, especially when exploring unfamiliar terminology. It should be treated as a starting point rather than an authority on disciplinary vocabulary.
Compare suggested terms with actual journal articles, database indexing, scholarly usage, and input from proficient speakers when available. The goal is not to produce an elegant translation. The goal is to retrieve the relevant literature.
Do not translate unnecessarily large search strings
Complex search strategies can contain dozens or hundreds of terms. Translating every free-text variant into several languages may produce an enormous query that is difficult to validate and impossible to use in simpler regional interfaces.
Instead, identify where translation is likely to add retrieval. Some concepts may already be captured well through English metadata or controlled vocabulary, while another concept needs a carefully constructed target-language term set.
Watch Out
A translated search strategy can look impressively comprehensive while retrieving poorly if the terminology was never validated against actual publications. Test translated terms using known relevant records rather than judging them by linguistic plausibility alone.
04 · A Practical Example
How to translate a search without merely translating a sentence
Hypothetical Example
Searching a national database for studies of student engagement
A researcher has an English search strategy for student engagement in higher education and wants to search a national database containing substantial literature in another language.
Inspect the database The researcher checks which fields are searchable, whether English metadata are common, whether controlled vocabulary is available, and whether original-language terms are necessary.
Define the concept Instead of sending the phrase “student engagement” through a translator and stopping there, the researcher clarifies which dimensions of engagement fall within the review.
Collect candidate terms Translations are generated with help from a proficient speaker and terminology observed in known relevant papers, database subject headings, and local scholarly publications.
Test known records The researcher checks whether those terms retrieve relevant articles already known to exist and identifies additional terminology from the retrieved records.
Refine the query Terms that retrieve useful literature are retained, while misleading or unproductive translations are revised or removed.
The result is not simply a translated English query. It is a search strategy adapted to the vocabulary and retrieval behavior of the target evidence source.
07 · A Quick Checklist
Before translating your search terms
Check whether you have:
Determined whether the target database already provides searchable English translations, abstracts, keywords, or controlled vocabulary.
Defined each search concept before attempting to translate individual terms.
Examined known relevant publications for terminology actually used by researchers in the target language.
Considered synonyms, abbreviations, spelling variants, regional usage, and discipline-specific terminology.
Checked whether original script, transliteration, or both are searchable in the target database.
Obtained input from someone with suitable language and subject knowledge when the terminology is consequential or ambiguous.
Tested translated terms against known relevant records rather than assuming that plausible translations work.
Documented translated terminology and database-specific adaptations when reproducibility matters.
