Using termbases
Use termbases to improve translations
Overview
Termbases are a critical part of modern translation workflows. They allow for consistency of translation across multiple languages.
CaptionHub includes built-in termbase support to help maintain consistent terminology across translations. You need to upload your termbase to CaptionHub, and then attach it to a project. Once attached to a project, the termbase will be automatically applied during machine translation when using Amazon Translate or DeepL. Matching terms will also be highlighted in the editor for linguists and reviewers.
CaptionHub applies termbases using language-aware matching logic. In most cases, termbases will still apply even if the project locale differs from the exact locale used in the CSV, as long as the base language matches. For example, an EN or EN-GB termbase can also be applied to English (US) projects, and vice versa.
There are two exceptions to this behaviour:
- CaptionHub will not apply termbases across different writing systems. For example, a Chinese (Traditional) termbase will not be applied to a Chinese (Simplified) project.
- If your termbase includes multiple locale-specific variants of the same language (for example both
EN-GBandEN-US), CaptionHub will prioritise the column that most closely matches the project locale.
Prepare your termbase for CaptionHub
CaptionHub supports termbases uploaded in CSV format. Your CSV needs to have at least two columns. Each column header must use the supported language code for that language. For example, English (Global) should be EN, and English (United States) should be EN-US or EN_US. Entries with empty cells will be ignored.
Therefore, a termbase with English (United Kingdom), Dutch and French would look like this:
EN-GB | NL | FR |
CaptionHub | CaptionHub | CaptionHub |
computer | ordinateur | |
email | e-mail | courriel |
You can continue to add extra columns to the CSV for multilingual termbases.
Creating a termbase
- Navigate to the Termbases page from the sidebar navigation
- Click New termbase
- Give your termbase a name and click Create termbase
- When your termbase is created, you need to click on the Upload termbase CSV button
- Locate the CSV file you prepared earlier, and click on the Import button
Attaching a termbase to a project
- From the Project page, click on the drop down More actions ⋯ menu
- Choose Attach termbases
- Select the termbase(s) you'd like to attach to your project, and click Done
When using Amazon Translate or DeepL, or when creating blank translations, CaptionHub will automatically apply the attached termbase where appropriate.
Automatically attaching termbases to projects
You can consistently attach a termbase to your project(s) by using an Automation template. The termbase needs to be attached to the template via the Project tab in the template configuration. For more details on setting up automation templates, please refer to this guide here.
It’s also possible to attach termbases by default by using the Mark as default action on the termbases page. In order to do this, click the kebab menu on a termbase in the termbases list (or the dropdown cog on the termbase edit page) and click the Mark as default option. Marking a termbase as a team default will mean it gets attached to every new project that’s created.
Marks as default action on the termbase edit pageGuidance for using termbases
- Termbases can be no larger than 10MB, and the maximum entry size is 200 bytes.
- Try to keep your custom terminology minimal. Only include words which you want to control and which are completely unambiguous. Only include terms where a single translation should always be enforced. Ideally, limit the list to proper names, like brand names and product names.
- Custom terminologies are case-sensitive. If you need both capitalised and non-capitalised versions of a word to be included, you must include an entry for each version.
- Do not include different translations for the same source phrase (for example, entry #1-EN: Amazon, FR: Amazon; entry #2-EN: Amazon FR: Amazon).
- Some languages do not change the shape of a word based on sentence context. With these languages, applying a custom terminology is most likely to improve overall translation quality. However, some languages do have extensive word shape changes. We do not recommend applying the feature to those languages, but we do not restrict you from doing so.
- If your workflow requires region-specific terminology differences (for example, English UK vs English US), we recommend including separate locale-specific columns in your CSV. CaptionHub will automatically prioritise the closest matching locale for the project language.
Troubleshooting
Error message | Solution |
The CSV file contains languages we don’t recognise | Please check that each column headers matches a supported language. The format should be CODE_TERRITORY, for example EN-US for English (United States), EN for English (Global), etc. |
Important: How termbases are applied
Termbases are not applied as a strict one-to-one “find and replace” system. Instead, CaptionHub incorporates termbase entries into the translation process while taking into account the grammatical and linguistic context of each sentence.
As a result, termbase entries may be adapted to fit the target language naturally. This can include changes to word form, agreement, or sentence structure.
In some situations, a term may therefore:
- Appear slightly differently from the original termbase entry
- Not be applied where a direct word-for-word substitution would produce unnatural or grammatically incorrect output
This behaviour is designed to preserve overall translation quality, readability, and fluency.