Language assets

Terminology extraction and termbase

Find bilingual term candidates, review them, and link approved terms to reusable glossary entries.

Terminology extraction proposes bilingual term candidates from your document. Review their meaning, choose a compatible glossary or create one, and approve the terms worth keeping. Approval into a glossary records a link to the reusable entry. The extraction record keeps the original candidate while showing the linked entry and its current translation. Available from Freelance upward.

How to use it

1

Open Language assets → Terminology

Upload a document here, or choose Extract terms beside a completed translation in Documents, including a PDF. Start extraction to identify term candidates: the vocabulary that carries the document's meaning, not every noun.

2

Review the candidates

Each candidate shows the source term, the proposed translation, and how often it appears. Approve what is right, reject what is not.

3

Let approved terms work for you

Choose a glossary with the same language pair, or select New glossary to create one during review. Approve the candidates you want to add, then choose Download glossary CSV. Different translations of the same source are preserved as alternatives. Review their usage labels in the glossary.

Review context and existing entries

Choose Review context beside an extraction candidate. Source context shows up to three bounded excerpts containing the term. New extractions retain these excerpts, including terms extracted from uploaded files. Older document extractions can use saved source segments; an older upload without stored context needs to be extracted again.

Select one of your glossaries with the same language pair to see existing source entries. The review distinguishes an existing bilingual pair from a different translation. Approving an existing pair reuses that entry; approving a different target preserves alternatives. Related extraction candidates group the same source wording, ignoring capitalization and outer spaces, within your termbase and language pair. The lists show up to 20 entries each. This is literal grouping, not automatic merging of inflected or semantically similar terms.

Review context does not approve, modify or delete anything. Close it and use the candidate’s edit, approve or reject controls. The glossary selected for comparison does not change the approval destination in the extraction list. Source excerpts are evidence of occurrence, not proof that the proposed target is correct.

Extraction proposes; the reviewer decides

Extraction is a starting list, not an approved glossary. It finds possible term pairs in material you have already translated and shows their frequency. Review each candidate for meaning and wording; rejected candidates are discarded. With Auto-approve enabled in Settings, candidates enter the approved terminology library immediately; they still need to be promoted to a glossary before that glossary can be used for translation.