Glossaries vs. Translation Memories
Glossaries and translation memories are both tools for consistency, but they operate at different levels and serve different purposes. Understanding the distinction helps you use each one effectively.
The Fundamental Difference#
A glossary is term-level and prescriptive. It defines how specific terms should be translated. "This word should always be translated as this." Glossaries represent deliberate decisions made by your team about terminology.
A translation memory (TM) is sentence-level and descriptive. It records how sentences were actually translated in past projects. "This sentence was translated as this before." TMs are built from completed work and grow automatically over time.
Think of it this way: a glossary is a dictionary your team writes. A translation memory is a library of your team's past work.
When to Use a Glossary#
Glossaries are essential when terminology must be consistent across translators, documents, and projects. Common scenarios:
Theological terminology. Words like "atonement," "sanctification," "redemption," and "covenant" carry specific meaning and often have multiple possible translations in other languages. Your glossary records which rendering your organization has chosen, ensuring every translator and every AI suggestion uses the same term.
Biblical proper names. Transliteration conventions vary by organization and target language. A glossary ensures "Jerusalem," "Moses," and "Paul" are always rendered the same way, following your organization's conventions.
Organization-specific terms. Program names, course titles, ministry names, and branded terms that should not vary. If your curriculum is called "Foundations of Faith" and you have decided on a specific translation for that title, the glossary enforces it.
Any term where inconsistency would cause confusion. If a reader encounters two different words for the same concept across chapters, it creates confusion about whether the author is referring to two different things. Glossaries prevent this.
Tip: Start your glossary before you begin translating. Even a small glossary of 20-30 key terms will noticeably improve consistency in AI suggestions from the first segment onward.
When to Use a Translation Memory#
Translation memories become valuable when your organization translates related content over time. Common scenarios:
Long-running translation programs. If you translate a 12-part curriculum series, later volumes will share phrases, standard instructions, and repeated content with earlier ones. A TM built from the first few volumes accelerates work on the rest.
Series of books or related documents. Books by the same author, articles in a publication series, or training modules in a program often reuse similar language. The TM captures how you translated it before.
Training materials with standard sections. Many training documents share introductory paragraphs, disclaimers, copyright notices, or methodology descriptions. Once translated, these appear as near-exact TM matches in future documents.
Building organizational knowledge. Every project your team completes can feed into the TM. Over months and years, the TM becomes a comprehensive record of how your organization translates, making every future project faster.
How They Work Together#
Your glossary and TM serve complementary roles when the AI translates a segment:
-
Glossary terms are treated as mandatory. If the segment contains a glossary term, the AI uses the specified translation. This is non-negotiable for confirmed terms.
-
TM matches are treated as helpful references. If the TM contains a similar sentence, the AI uses it as a guide for phrasing, style, and structure -- but adapts it to fit the current context.
The AI combines both: it uses your glossary's exact terms and your TM's sentence-level patterns to produce a suggestion that is terminologically consistent and stylistically aligned with your past work.
Consider a sentence about "the new covenant." Your glossary ensures "covenant" is translated with the right term. Your TM, containing a similar sentence from a previous project, helps the AI construct the full sentence in a style that matches how your team has handled similar language before.
Building Knowledge Over Time#
Both glossaries and TMs become more valuable as they grow:
-
Glossaries start small and grow as you encounter new terms that need standardization. A mature glossary for a language pair might contain hundreds of terms covering theological vocabulary, proper names, and organizational terminology.
-
Translation memories grow automatically. Every project you complete can contribute its translated segments to your TM. A TM that starts empty after your first project might contain thousands of entries after a year of work.
The investment is cumulative. An organization that has been using Lectio for a year, with a well-maintained glossary and a TM built from dozens of completed projects, will see dramatically faster and more consistent AI suggestions than one starting from scratch. Every term you add and every project you complete makes the next project easier.
Practical Guidelines#
| Glossary | Translation Memory | |
|---|---|---|
| Granularity | Individual terms and short phrases | Full sentences and paragraphs |
| Source | Human decisions (curated) | Completed translations (automatic) |
| AI treatment | Mandatory (confirmed terms) | Advisory (reference material) |
| Best started | Before translating | After completing your first project |
| Maintenance | Active curation -- add, review, confirm | Mostly automatic -- review periodically for quality |
| Shared across | All projects in the same language pair | All projects in the same language pair |
Note: Glossaries and TMs are language-pair specific. A glossary for English-to-Thai and a TM for English-to-Thai do not interact with English-to-French resources.
On the source side, English, English (American), and English (British) count as one language: projects in any of them share the same glossaries, translation memories, and rules, and they use a single language pair of your plan. The variant you pick still tells the AI how to read dialect-specific wording, such as whether 04/05 means April 5 or 4 May.
For details on importing and exporting glossary and TM data, see Glossary and TM File Formats.