Documentation Center

AI Assistant Reference

Learn all about the options and settings in AI Assistant.

General tab

Connections group box: use this group box to add, edit and delete your LLM connections. With OpenAI provider, you have the choice of gpt-4 and gpt-3.5 models depending on your needs for performance, cost and accuracy. gpt-4 models tend to be more expensive but also more powerful. We recommend the 4o model for best price/performance ratio. Use 3.5 for more cost-effective connections. For any LLM usage, we recommend that you monitor the usage and billing in your LLM provider's account, such as OpenAI.

Search options group box

OptionExplanation
Automatically search for translation suggestionsGoverns the behaviour of AI Assistant in its translation companion capacity when working on documents in the Studio Editor. Use this setting with care. When checked, AI Assistant will always connect to the LLM to look up the current segment when you get to it. This can be useful when you need suggestions continuously while working through your document, but will incur higher cost as the current prompt will be sent to the LLM for each segment. The Ignore translated segments option is useful for managing cost when automatically searching for suggestions. Any segment whose status is Translated, Translation Approved/Rejected or Signed Off/Sign-off rejected will not trigger a lookup in the LLM.
Ignore locked segmentsSimilarly to Ignore translated segments, this option is useful for managing cost when automatically searching for suggestions. Any locked segment (no matter its status) will not trigger a lookup in the LLM.
Enable terminology aware translation suggestionsThis option enhances translation suggestions by incorporating terminology context with the prompt when requesting translations from the LLM. By enabling this feature, the LLM can refine the generated translation suggestions using translated terms from your termbase, leading to more accurate and contextually aware translations. This integration of terminology context ensures precision and consistency in your translated content. For an example on how to use it, see Using AI Assistant to your Project as a Terminology-Aware Translation Provider.
Include target segment with the translation requestEnable this option if you are explicitly making reference to the existing translation in the prompt and therefore should be included in the translation request, for example, if the prompt is using the existing translation as context to provide the new translation. This option should also be used for review-oriented prompts such as "rewrite this translation using gender neutral style".
Include tags in the translation suggestionsThis option governs the behaviour for source content with tags or formatting. When active, AI Assistant will provide translations that will contain any tags or formatting in the source and transfer them to the target segment. This is the recommended default as it will ensure that tags and formatting are also present in the target text. However, if you wish to work just with plain text without formatting or tags, then you can switch this option off. This can sometimes be useful in heavily tagged or otherwise formatted text to clean out that formatting in the target and reapply it manually. You can also decide how tags should look in the AI Assistant view part. The options here are the same as for Studio's Editor: Tag ID displays the tag's ID, hide tags will hide the tags, no tag text shows tag placeholders only, partial tag text (default) will display the main tag information and full tag text will display all the text on a tag.
Use In-Process CachingSelect this option to reuse translations already obtained from the LLM during the current Trados Studio session. This is useful to save cost and to send any request to the LLM only once. After making changes such as changing or adding new terms to refine translation suggestions, it can be useful to refresh the cache so that any request to the LLM is sent again. You can use the Reset cache option to achieve this. This will emtpy the cache and will enforce new lookups.

Search results group box

Assign Match value:
This setting governs what match value the AI Assistant should apply to any translated segments both in batch processing (Pre-Translate Files task) and interactive translation. Behind the scenes, AI Assistant works in a hybrid fashion. On the one hand, it provides results with an NMT (Neural Machine Translation) origin, on the other, it assigns a fuzzy match value like a TM would. This hybrid operation provides a few benefits:
  • You can easily filter for any segments processed by AI Assistant. For instance, if you assign a match value of, say, 85, then you will be able to filter for that match value in Advanced Display Filter.
  • This setting also governs 'double pretranslation' where you first translate using NMT and then the OpenAI provider or other LLM to refine the translations in batch processing. If you assign a value of 90, for example, then you can use a Pre-translate value of 90 or lower in the pre-translate settings to make sure that the OpenAI provider can overwrite existing NMT translations.
Display comparison differences in the Translation Suggestions:
This setting governs the behavior in the AI Assistant when used as a translation companion. When checked, any differences between the original target segment and the refined AI translation suggestion will be displayed in tracked changes formatting. Switch this setting off to display the translation suggestions as is. This can be useful when the changes are too extensive, which can be confusing.

Sign Up for an OpenAI Account

Click on this link to sign up for an OpenAI account quickly and easily. A paid OpenAI or AzureOpenAI subscription is required to use AI Assistant in Trados Studio.

Prompts tab

Use this tab to edit, create or delete existing prompts. See Working with Prompts

Customer Experience tab

Use this tab to decide if you want to join the customer experience program and help improve AI Assistant. We don't collect any personally identifiable information nor any of your content. We recommend that you join the customer experience program, as this allows us to idenfify usage patterns in our user base, which in turn can help us refine existing or create new functionality in future Trados Studio versions.