How to Keep Names and Terminology Consistent Across a Translated Book

Names and terminology stay consistent when one approved record controls every chapter. Build that record before translation, separate names from domain terms and style decisions, give each entry one approved target form, record rejected variants, and search the complete translated file after every major revision.

For a book-length project, use four artifacts:

1. A terminology glossary for concepts and repeated phrases.
2. A character or entity bible for people, places, organizations, and relationships.
3. A do-not-translate list for codes, brands, URLs, formulas, and intentional source forms.
4. A variant log that records wrong forms found during review.

Download the book terminology and character-bible template. The example rows are deliberately fictional, so you can see how approved and rejected forms work without mistaking them for a real project.

Treat consistency as identity control, not word replacement

The hard problem is not choosing a translation once. It is making sure every later occurrence refers to the same person or concept, even when capitalization, inflection, abbreviations, scripts, or context change.

Separate the control list into three layers:

This prevents two bad fixes. First, a reviewer does not replace every matching string when two concepts happen to share a source word. Second, a harmless grammatical inflection is not treated as drift merely because it differs from the glossary's citation form.

The European Commission's 2024 translation quality evaluation information pack defines terminology errors in relation to accepted domain usage, a term base, or a supplied reference document. It also includes names, places, and other named entities in accuracy review. That distinction maps well to the identity and concept layers above.

Build the glossary before translating the full book

Start with the material most likely to repeat or cause confusion:

• the table of contents and chapter titles;
• the first two representative chapters;
• indexes, glossaries, cast lists, and appendices;
• bold or italic definitions;
• recurring capitalized phrases and acronyms;
• character, place, organization, product, and series names;
• subject-specific terms; and
• words the author intentionally uses in a nonstandard way.

Do not aim to list every noun. Prioritize entries where inconsistency changes identity, meaning, navigation, or reader trust.

Each entry should answer these questions:

The European Commission's current translation resources put terminology databases and style guides side by side. A project glossary should do the same job at a smaller scale: record both what a concept is called and how the approved form is written.

Keep a separate character and entity bible

A glossary entry is often too small for fiction, biography, history, or narrative nonfiction. A character bible should record:

• full name and short forms;
• native-script form and approved transliteration;
• titles, ranks, honorifics, and pronouns;
• relationships and aliases;
• whether the name's meaning matters to the story;
• dialogue or register notes; and
• the chapters in which a change is intentional.

For organizations and places, add abbreviations, historical names, local names, and whether the project translates or transliterates the form.

Do not silently localize every meaningful name. A fictional name may carry a joke that needs adaptation, while a real person's name may have an established target-language spelling. Record the decision and its evidence. The goal is not one universal naming rule; it is one explicit rule per entity.

Use a do-not-translate list narrowly

A do-not-translate list is for content that must remain exact, such as:

• URLs, email addresses, file paths, and code;
• product and company names that have no localized form;
• model numbers, warning codes, legal identifiers, and formulas;
• bibliographic strings that must match a cited edition; and
• intentional foreign-language phrases.

Do not put a term on the list merely because its translation is difficult. That hides a terminology decision instead of making one.

The W3C's Internationalization Tag Set 2.0 includes a mechanism for marking content as translatable or not translatable. A spreadsheet glossary is less sophisticated, but the principle is the same: the instruction has to travel with the content rather than live only in a reviewer's memory.

A deliberately drifted example

Imagine a fictional fantasy manuscript with these source entities:

After translation, full-file search finds:

• two spellings of the North Gate name;
• "the keeper" in one dialogue scene;
• a literal translation of ashglass in chapter 3 and a transliteration in chapter 11; and
• the approved coined term in the other 22 occurrences.

The right fix is not three isolated edits. Add the wrong forms to the rejected-variant column, search every occurrence, inspect ambiguous hits in context, update the affected chapters, and rerun the search on the final export.

Change the method for fiction, nonfiction, and technical books

For research material, the academic paper translation workflow adds checks for equations, citations, figures, and domain review. For long-form production decisions, use the complete book translation guide.

Apply the glossary during translation without freezing the language

A glossary should control identity and meaning, not force every sentence into the same surface form.

During translation:

1. Freeze version 1 of the glossary and source file.
2. Translate a representative chapter.
3. Review approved terms in full sentences.
4. Correct decisions that are unnatural or ambiguous before scaling.
5. Record every approved change with its affected chapters.
6. Translate the remaining book using the new glossary version.
7. Do not overwrite an older decision without preserving the rejected form for QA.

In inflected languages, the glossary's approved target is often a lemma or citation form. The translation can legitimately use grammatical variants. Document the allowed pattern so automated search does not create false positives.

For mixed scripts, store both native-script and transliterated forms. Decide whether readers need one, the other, or both on first mention. Use a language tag or locale only when it adds a real distinction; the W3C's language-tag guidance recommends keeping tags as short as possible unless script, region, or another subtag is necessary.

Find terminology drift after translation

Run the review against the complete target file, not a sample alone.

1. Search approved and rejected forms

For each high-risk entry, record:

• count of the approved form;
• count and locations of every rejected form;
• case variants;
• source form left untranslated; and
• unexplained zero counts.

A zero count may mean the concept was paraphrased correctly, omitted, or missed by extraction. Inspect the source location before deciding.

2. Review chapter boundaries

Terminology often drifts when work is split into batches. Compare the last section of one chapter with the first section of the next and audit every handoff between translators, models, or editing sessions.

3. Search abbreviations and expansions together

Confirm that the first-use expansion and later acronym refer to the same concept. Check plural forms and translated versus untranslated abbreviations.

4. Recheck the exported file

A correct editing file can produce a wrong EPUB, PDF, or DOCX when styles, fonts, metadata, or conversion rules change. Search the deliverable readers will receive.

5. Turn every confirmed error into a reusable rule

Add the wrong form, affected locations, root cause, and correction to the variant log. The next review should start with known failure patterns instead of rediscovering them.

Decide when to use software and when to use a reviewer

Search and terminology tools are good at finding repeated forms. They cannot always decide whether a variation is grammatical, stylistic, or wrong.

Use automated checks for:

• exact approved and rejected strings;
• capitalization and punctuation variants;
• source-language residues;
• acronyms, model numbers, and codes; and
• changes between glossary versions.

Use a qualified reviewer for:

• ambiguous concepts;
• established names and domain conventions;
• literary voice, wordplay, and intentional variation;
• sensitive or high-stakes terminology; and
• any correction that changes meaning beyond one term.

If you want a complete-file first draft, BookTranslator handles books and long documents rather than requiring manual copy-and-paste. Its current translation form offers an optional Auto glossary for eligible non-PDF tasks, collecting recurring names, places, and specialized terms. Keep your project glossary anyway: it records human decisions, rejected variants, and the acceptance evidence needed after translation.