Termbase Management

A termbase (TB) keeps one concept rendered the same way across every file. In Transbox a term is a multi-language text entry: the same concept expressed in each language, importable and exportable in the TBX standard.

Managing Termbases

Open Corpus Management → Termbase.

List and actions

Columns: index, name, languages, entry count, creation time. Filter by name; select rows to merge.

Row actions:

• Details: browse and search entries
• Import: add terms
• Export: TBX / Excel / split export by file count / split export by entry count
• More: rename / clear / re-index / delete

The Term recognition entry sits at the top right (see below).

Import

Export

• Export TBX: standard format for other CAT tools
• Export Excel: for sharing and manual revision
• Split export by file count / entry count: cut the termbase into several files

Browsing and Editing

• One column per language contained in the termbase
• Keyword search with an optional language filter; matches highlighted
• Double-click a row (or Edit) to change its texts; Delete removes one entry
• Results are paged

Entry structure

One-click dedupe

Shares the memory cleanup rules:

• Ignore case
• Delete entries containing empty text
• Delete an entry when any two of its languages are identical
• When a selected language is duplicated, keep one entry

Term Recognition

Click Term recognition at the top right of the termbase list:

1. Choose the recognition language
2. Paste or type the text to analyze
3. Click Recognize: terms found in the local termbases are highlighted
4. Click a highlighted term to see its text in every language

Useful before translating, or when checking whether an external text already has established terminology. The scope is all local termbases.

Using Termbases in Translation

• Corpus panel: the editor's right-hand Corpus tab shows term matches for the current segment
• Add term: the Add term tab turns the current source/target pair into a term entry
• LLM and Agent translation: matched terms (up to 10 entries whose source and target differ) are injected into the prompt, instructing the model to use them
• QA: terminology inconsistencies are reported as QA issues when the target does not use the termbase rendering

Best Practices

• One termbase per domain: medical, legal and IT kept apart avoids homonym conflicts
• One rendering per term: keep a single standard translation inside a termbase
• Verify before import: an LLM will follow a wrong term literally
• Split responsibilities: termbases own words, memories own sentences

Notes

1. Capacity: the entry cap per termbase depends on the plan
2. Languages come from data: importing a file defines the language columns
3. Deleting and clearing are irreversible
4. Search: local tokenized full-text index

FAQ

Q: What is TBX?

A: TBX (TermBase eXchange) is the standard exchange format for terminology data, supported by SDL MultiTerm, memoQ and others, which makes it the portable choice.

Q: How is a termbase different from a memory?

A: A termbase stores words or phrases for consistent rendering; a memory stores full source–target sentence pairs for reuse. Both participate in matching but do different jobs.

Q: Why are there no definition or part-of-speech fields?

A: Entries keep only the per-language term text. Keep extra metadata in your source Excel/TBX, or write it into the term text itself.

Q: Terms are not suggested in the editor.

A: Check that the termbase covers the segment's source and target languages, and that the term text matches the source exactly (case and spaces included). After large imports, run Re-index if search looks incomplete.

Q: Can I share a termbase with colleagues?

A: Yes — export TBX or Excel and let them import it, or share a full data backup.

Q: What about terminology conflicts (two renderings for one source)?

A: Prefer separate termbases per client/domain, or dedupe by language so only one standard rendering remains.