What is an MT-Optimized Glossary?
An MT-Optimized Glossary is used for Glossary Term Insertion in machine translations.
Unlike human linguists, machine translation engines cannot distinguish which glossary terms should be used in a particular context only. Therefore, we recommend optimizing your glossary for machine translation before integrating it with an MT Profile.
Only glossary entries that perform well with machine translation and can easily be recognized by MT engines (such as nouns and noun phrases) will be used by the MT engine.
For best results, the glossary should include only nouns and noun phrases when it is used for insertion in MT. Other entries should be removed to reduce the risk of mistranslations.
This can easily be achieved without impacting your existing glossaries for human translation, by creating a separate, MT-Optimized Glossary for use in machine translations.
Smartling has automated this process, so no manual effort is required.
Only one MT-Optimized Glossary can be created for your Smartling account.
How to create an MT-Optimized Glossary
- From your Account Settings, click Linguistic Assets > Glossaries
- Click Optimize Glossary for MT
- Select one or more glossaries
- As each MT Profile can only use one glossary, the entries from all selected glossaries will be combined and converted to a single, MT-Optimized Glossary.
- Configure deletion severity level to one of the following:
- Strict (default, recommended)
- Allow plurals: Plurals are not removed, and multi-word terms with 5 or more words are removed
- Allow all parts of speech: Plurals are removed, and multi-word terms with 5 or more words are removed
- Most Liberal: Plurals are not removed, and multi-word terms with 7 or more words are removed
- To create the MT-Optimized Glossary, click Confirm.
All selected glossaries will then be combined into a single, account-wide glossary for use in MT, called "MT Optimized Glossary".
The MT-Optimized Glossary is created as a separate, stand-alone glossary. The original glossaries remain intact, and no terms are removed from them.
The glossary conversion process must be completed before you can add it to an MT Profile. This may take several minutes. A success message will be displayed when the MT-Optimized Glossary is ready.
Click View Details on the success message to view the conversion details:
- Percentage of entries removed
- Number of removed entries which were not suitable for machine translation
Which entries are removed from an MT-Optimized Glossary?
During the conversion process, entries which are not suitable for machine translation are removed to ensure high-quality MT output, including:
- Any duplicate terms or plural forms
- Glossary terms including four words or more, or containing HTML tags (such as slogans, longer phrases or even full sentences)
For best results, the MT-Optimized Glossary includes only nouns and noun phrases. Plurals and other parts of speech, such as verbs and adjectives, can be highly inflected in many languages, and MT engines may not correctly inflect Glossary terms that are verbs or adjectives.
The glossary conversion supports all glossary languages and all languages that the MT profile supports. The optimization looks at entries in two of the following languages: English, Korean, Japanese, Chinese, Portuguese, Russian, French, Italian, German and Spanish. If an entry does not meet appropriate standards to be used in machine translation for one of the two languages checked, the whole entry gets removed.
Entries that are marked as DNT (Do Not Translate) and translated entries can both be inserted in machine translations.
Deletion Severity
You can control how strict terms are removed from the MT-Optimized Glossary. Strict sets the conversion to keep only nouns and noun phrases. You can choose to include other parts of speech or plurals, or both. When changing the deletion severity to anything other than strict, consider whether this glossary will be used in a workflow with human edit and review.
Tip: For a CSV file detailing the discard reasons for all terms that were removed, reach out to your Smartling Customer Success Manager.
How to update an MT-Optimized Glossary
Once an MT-Optimized Glossary is created, it cannot be archived and terms cannot be edited or removed. You will see an MT OPTIMIZED lock to indicate that this glossary cannot be edited.
MT-optimized glossaries are automatically updated when the source glossaries change, such as:
- Any entries which have been deleted or archived from the source glossaries will be removed from the MT-Optimized Glossary.
- Any new entries will be added to the MT-Optimized Glossary.
- Updated terms will be overwritten in the MT-Optimized Glossary.
The automatic update can be seen in the date modified column, as updated by "Automatic Glossary Sync".
Note: The automatic update happens once in 24-hours.
Adding an MT-Optimized Glossary to an MT Profile
For insertion in machine translations, simply add it to an MT Profile.
Go to Account Settings > Machine Translation Settings > Profiles and select the "MT Optimized Glossary" when enabling Glossary Insertion for an MT Profile.
The MT-Optimized Glossary will then be used for all machine translation workflows and account-wide machine translation options where the associated MT Profile is used, including:
✅ Smartling Translate
✅ MT suggestions or autopopulate options in the CAT Tool
✅ Zendesk Support Plugin
✅ Site-wide machine translation for the GDN
✅ Smartling’s Machine Translation API
If you would like to use the MT-Optimized Glossary for workflows using human translation, you can assign it to the relevant Linguistic Package(s).