Skip to main content

XTM Cloud 26.2

Creating a project: Machine translation

This section provides information about optional Machine translation settings that you can make while creating a project in XTM Cloud.

XTM Cloud is integrated with a wide selection of Machine Translation (MT) engines. Only MT engines that have been set up by your XTM Cloud Administrator for the entire system, or by the Project Manager at customer level, are visible and available for selection during project creation.

The supported MT engines are:

  • Acolad.

  • Amazon Translate.

  • CrossLang.

  • Custom MT.

  • DeepL.

  • Google Translate.

  • Intento.

  • KantanAI.

  • Language Weaver.

  • Lingo24.

  • Lionbridge GeoFluent.

  • Microsoft Translator.

  • Omniscien Technologies Language Studio.

  • Oneliner.

  • OpenAI GPT.

  • SAP Translation Hub.

  • SmartMATE.

  • SYSTRAN.

In the Creating project > Machine translation section, you can activate/deactivate all MT engines that have been integrated with your instance by your XTM Cloud Administrator. XTM Cloud will send a request to all MT engines that you have enabled for a project, for all language pairs in the project. Your XTM Cloud Administrator can specify whether an MT request will be send out during project analysis, or on a segment-by-segment basis, in XTM Workbench, or both. XTM Cloud will never send a new request to the MT engine if a machine translation match for a segment has already been returned. This also applies when source files have been updated (continuous projects), for segments whose content and context (their previous and the following segment) have not changed.

Warning

Project reanalysis means that all source files are analyzed again. For this reason, machine translation requests will be sent again in accordance with the current project settings.

Language pair support can be limited by your MT engine provider or by the additional MT settings, such as models or profiles. For the following MT engines, it is also possible to map MT engine usage for particular language pair(s):

  • CrossLang.

  • DeepL.

  • Google Translate – AutoML, including NMT.

  • Intento.

  • Language Weaver.

  • OpenAI GPT.

  • Omniscien Technologies.

  • SAP Translation Hub.

  • SYSTRAN.

  • Any MT integration which uses Custom MT.

In the case of the AWS Amazon Translate MT engine, you can only map its use for terminology for specific language pair(s).

If the same language is sent to more than one MT engine, XTM Cloud will send a translation request to each of them. In XTM Workbench, the MT engine results for the relevant segment(s) are displayed in the Matches docked panel. Each result is labeled with the relevant MT engine name. The order in which the results are displayed depends on the order in which the MT engines respond. The Populate target with match setting will be applied to the first MT match on the Matches list.

Important

Ensure that you always use an MT engine which is compliant with your security and data-processing requirements.

Important

If a project is subcontracted, your machine translation (MT) settings are inherited by your subcontractor (LSP) and their Linguists.

Neither LSP Project Managers nor LSP Linguists can change these settings.

Table 68. Creating a project: Machine translation settings

Setting

Description

Mandatory/

Optional

Type

Valid values

Custom MT settings

Use [your custom MT's name]

Use this setting to specify whether this custom MT is to be used for the project.

The custom MT name has been specified by your XTM Cloud Administrator. For details, see Connecting a Custom Machine Translation engine.

Optional

Checkbox

Valid values:

  • Cleared – to disable the selected custom MT engine for the project.

  • Selected – to enable the selected custom MT engine for the project.

    When you select this option, the Model for language pair setting is displayed.

The default is specified in system settings or customer settings.

Note

The customer settings take priority over system settings.

Model for language pair

This setting is displayed if you have selected the Use [your custom MT's name] checkbox.

When you select the Choose model button, the Model for language pair popup is displayed. Use it to enable your custom MT for the required project language pairs.

Optional

Button

When you select the Choose model button, the Model for language pair popup is displayed.

In the Model for language pair popup, select the appropriate option for each target language. Then, at the bottom of the Model for language pair popup, select OK. The options are:

  • Do not match for this language – your custom MT engine is deactivated for the relevant project target language.

  • Use custom MT – your custom MT engine is activated for the project target language.

The default option is Do not match for this language, but it can be changed to Use custom MT by XTM Support.

To request support

Visit our Support Portal.

DeepL MT settings

Use DeepL Translator

Use this setting to specify whether the DeepL Translator MT engine is to be used for the project.

Note

The Use DeepL Translator entry in the table serves as an example for MT applied during a project creation.

Optional

Checkbox

Valid values:

  • Selected.

    When you select this option, the Advanced options setting is displayed.

  • Cleared

The default is specified in system settings or customer settings.

Note

The customer settings take priority over system settings.

Glossaries and formality

Note

The Glossaries and formality setting is only available for DeepL Translator

If you select this button, you can make these settings for each target language in the project:

  • Use MT for this language.

  • Glossary.

  • Formality.

Optional

Button

Advanced options values that are available for every target language:

  • Use MT for this language toggle switch:

    • Disabled.

    • Enabled.

      The Glossary and Formality dropdown lists is activated.

  • Glossary dropdown:

    • Not available — when DeepL Translator does not support glossaries for this target language.

    • Choose... — to select a glossary available for this target language.

  • Formality dropdown:

    • Not supported — when DeepL Translator does not support a formality setting for this target language.

    • Default (default).

    • More.

    • Less.

Google MT settings

Use Google MT

Use this setting to specify whether Google MT is to be used for the project.

Optional

Checkbox

Valid values:

  • Selected.

    When you select the Use Google MT checkbox, the Use AutoML checkbox is displayed.

  • Cleared.

The default is specified in system settings or customer settings.

Note

The customer settings take priority over system settings.

Use AutoML

Use this setting to specify whether Google AutoML is to be used in the project.

Optional

Checkbox

Valid values:

  • Selected.

    When you select the Use AutoML checkbox, the Models and glossaries setting and the Advanced options button are displayed.

  • Cleared.

Models and glossaries

You can select a model and glossary for a particular language pair in the project.

Optional

Button and popup

When you select the Advanced options button, the Advanced options popup is displayed.

In the Advanced options popup, for each Target language, select one of the following options:

  • Use MT for this language toggle switch:

    • Disabled.

    • Enabled.

      The Model and Glossary dropdown lists are activated.

  • Model dropdown:

    • Not supported — when Google AutoML does not support a model for this target language.

    • Google NMT (default).

      When you select this option, you are not able to select a glossary for this target language.

    • A list of models available in your Google AutoML account.

  • Glossary dropdown:

    • Not available — when Google AutoML does not support glossaries for this target language.

    • Choose... — to select a glossary that is available for this target language.

OpenAI GPT settings

Use OpenAI GPT

If your XTM Cloud Administrator has connected OpenAI GPT to your XTM Cloud instance, you can decide whether to enable or disable it for the project.

Optional

Checkbox

Valid values:

  • Cleared – to disable OpenAI GPT for the project.

  • Selected – to enable OpenAI GPT for the project.

    When you select this option, the Matching for language pair setting is displayed.

The default is specified in system settings or customer settings.

Note

The customer settings take priority over system settings.

Matching for language pair

This setting is displayed if you have selected the Use OpenAI GPT checkbox.

In Matching for language pair, you can enable OpenAI GPT for each project language pair.

Optional

Button

Select the Choose button to the right of the Matching for language pair label. As a result, the Matching for language pair popup is displayed.

In the Matching for language pair popup, for each project language pair, you can set the toggle switch to:

  • Disabled (default).

    OpenAI GPT MT is NOT activated for the particular target language.

  • Enabled.

    OpenAI GPT MT is activated for the particular target language.

AI automation section

AI Translate & TQI settings

Business use case

The AI Translate & TQI feature helps you reduce translation costs and improve workflow efficiency by enabling Linguists to focus exclusively on translations that really require post-editing or human intervention. With AI Translate & TQI, you can benefit from context-aware translations in line with your terminology, translation memory, context, and Style guides. Additionally, the feature provides you with quality scores based on configurable thresholds, so you can base the optimization of your review process on actual data, in line with industry standards.

AI Translate & TQI can be combined with Intelligent Workflow for greater impact and significant time savings. For details, see Global Intelligent Workflow settings.

Use AI Translate & TQI

Note

This feature is based on AI Words: source words sent to the LLMs. Words from the translation memory (TM), terminology, and context segments are NOT counted as AI Words. For details about AI Words calculation and deduction for projects, see AI Words in XTM Cloud. If your XTM Cloud instance uses an XTM-provided LLM subscription, an AI Words allowance must also be purchased for your entire XTM Cloud instance. For more information, contact your account manager.

XTM uses ChatGPT and Gemini to run the AI Post-Edit feature. To check which models are currently used, see the Supported models section in https://xtm.ai.transifex.com/docs#section/LLM-Configuration.

Note

If you are using your own LLM keys, the size of the source text, terminology, TM records, context segments, and the Style guide instructions can influence the LLM usage costs. For details, refer to your LLM subscription rules.

This setting is only displayed if the AI Translate & TQI feature has been activated for your instance by your XTM Cloud Administrator. For details, see Global AI Translate & TQI settings.

The AI Translate & TQI feature automatically evaluates the quality and accuracy of translations generated by an LLM (Large Language Model). For every eligible segment, the feature compares translations generated by two or three different LLMs and evaluates their similarity. In the process, it analyzes the source segment, context (the previous and next segment), the three highest translation memory (TM) matches that were added or updated most recently, terminology, and the project's subject matter. The lower the similarity between these LLM-generated translations, the lower the quality evaluation score (percentage). The quality ranges can be configured by selecting the required values for the Unacceptable, Acceptable, and Good thresholds at global, customer, and project levels. Source text words processed by the AI Translate & TQI feature are referred to as AI Words. Words from the translation memory (TM), terminology, and context segments are NOT counted as AI Words.

Important

If the AI Translate & TQI feature is set to run in analysis and you use XTM Terminology, ensure that your XTM Cloud Administrator has enabled the Match terms in analysis setting (for details, see Terminology engines). If they haven't, as a user with the Project Manager role, select it yourself at customer level (for details, see Customer Editor – Settings). Otherwise, the AI Translate & TQI feature will not have access to terminology.

To improve the accuracy of LLM-generated translations, especially if the necessary terminology and translation memory (TM) records either do not exist or are very limited, we strongly recommend using a Style Guide. For details, see Style Guide in the table below.

The old Style Guide solution, created on the basis of the custom field (for details, see Creating a project: Custom fields), will be deprecated in the XTM Cloud version 26.2. We recommend that you start using the new Style Guide feature, in which your XTM Cloud Administrator creates multiple Style Guides globally for your instance and you, as a Project Managers can then select one of them for your project. For details, see Style Guide in the table below.

Important

The AI Translate & TQI feature will not process a segment for which a match type, selected in the Do not search for an MT/LLM match when one of the selected matches exists setting, has been found. For details, see the "General settings for machine and LLM-generated translations" section of this table.

If the Populate target with match setting for any translation memory match category is set either to Always, fixing inline differences or Only if inlines are the same, this TM match, if available, will automatically be inserted in the target segment, instead of the AI Translate & TQI translation. To ensure that an LLM-generated translation will take priority over TM matches and always be inserted in the appropriate target segment, automatically, ensure that the Populate target with match setting is set to Never. For details, see the "General settings for machine and LLM-generated translations" section of this table.

Optional

Checkbox

Valid values:

  • cleared – the AI Translate & TQI feature is deactivated for the project. If required, you and other Project Managers can then activate the feature in the Project Editor.

  • selected – the AI Translate & TQI feature is activated for the project. If required, you and other Project Managers can then deactivate the feature in the Project Editor. When you select the checkbox, these settings are displayed:

    • Matching for language pair.

    • AI Translate & TQI settings.

    • Confidence threshold.

The default is specified in system settings or customer settings.

Note

The customer settings take priority over system settings.

Matching for language pair

This setting is only displayed if you have selected the Use AI Translate & TQI checkbox.

You can specify the project target languages for which the AI Translate & TQI feature is enabled.

Optional

Button and popup

When you select the Choose button, the Use AI Translate & TQI popup is displayed.

In the Use AI Translate & TQI popup, select the appropriate setting for each target language checkbox. Then, at the bottom of the Use AI Translate & TQI popup, select Save. The checkbox setting options are:

  • cleared (default) – the AI Translate & TQI setting is deactivated for the project target language. If required, you and other Project Managers can then activate the feature in Project Editor.

  • selected – the AI Translate & TQI setting is activated for the project target language. If required, you and other Project Managers can then deactivate the feature in Project Editor.

Style Guide

This setting is only displayed if you have selected the Use AI Translate & TQI checkbox.

We strongly recommend that you use a Style Guide to improve the accuracy of LLM-generated translations, especially if the necessary terminology and translation memory (TM) records either do not exist or are very limited.

Important

Currently, it is not possible for XTM Cloud Project Managers to see the rules set for a particular Style Guide. To ensure that you select the correct Style Guide for your project, contact your XTM Cloud Administrator.

Currently, it is not possible to change this setting for an existing project in the Project Editor.

You can select one Style Guide for your project. General rules of that Style Guide will be applied for all target languages. Language-specific rules will only be applied to the relevant target language. Language-specific Style Guide rules that have been set for target languages that are not used in a project will be ignored. If there is a conflict between a general rule and a language-specific rule, the language-specific rule takes priority.

Optional

Dropdown

Select one of the active Style Guides created by your XTM Cloud Administrators for your instance. For details, see Style Guide tab.

AI Translate & TQI settings

You can specify which AI Translate & TQI thresholds are to be used for the project: the system default thresholds or project-specific settings.

This setting is only displayed if you have selected the Use AI Translate & TQI checkbox.

Mandatory if you have selected the Use AI Translate & TQI checkbox.

Dropdown

Select the appropriate option from the AI Translate & TQI settings dropdown:

  • System default (default) – the Confidence threshold section displays the system default values in read-only mode. These are the values set for the AI Translate & TQI feature at customer level (for details, see Customer Editor – Settings) or at global level, if customer-specific settings have not been made (for details, see Managing AI Translate & TQI at global level.

  • Project specific – the Confidence threshold section displays the system default values, but they can be edited, if required.

Confidence threshold

This setting is only displayed if you have selected the Use AI Translate & TQI checkbox.

Specify the ranges within which target segments are considered poor, medium, or high quality.

Optional

Text fields and sliders

Use the Confidence threshold section settings to specify the minimum quality score for acceptable or good translations:

  • Unacceptable – set a threshold below which segments are considered poor quality and need correction or post-editing by a Linguist. To do so, enter a numeric value in the text field. Alternatively, use the slider displayed above the Unacceptable label: the corresponding numeric value is displayed automatically in the text field. The Unacceptable quality range is marked in red on the slider. Target segments within this range are also marked in red in XTM Workbench. The default value is 0%.

  • Acceptable – set a threshold to create a range for medium-quality segments. These LLM-translated segments meet basic standards, but a Linguist might need to make minor improvements to them. To set the threshold, enter a numeric value in the text field. Alternatively, use the slider displayed above the Acceptable label: the corresponding numeric value is displayed automatically in the text field. The Acceptable quality range lies between the Unacceptable threshold, or 0% (if the Unacceptable threshold has not been set), and the Acceptable threshold value. It is marked in yellow on the slider. Target segments within this range are also marked in yellow in XTM Workbench. The default value is 0%.

  • Good – the setting value is fixed at 100%. Segments in the range between the Acceptable threshold, Unacceptable threshold (if Acceptable has not been set), or 0% (if neither the Acceptable nor Unacceptable threshold has been set), and 100% are considered high quality. They do not need corrections or post-editing by a Linguist. The Good quality range is marked in green on the slider. Target segments within this range are also marked in green in XTM Workbench.

    Important

    To ensure the best possible AI-generated translation quality, we recommend that you select the 95–100% Good quality range.

Intelligent Workflow section

Business use case

The Intelligent Workflow feature accelerates time to market, by reducing the overall workflow turnaround times and automating workflow decisions based on actual quality data. I also helps reduce translation costs and ensures efficiency in the translation workflow. Intelligent Workflow uses the AI Translate & TQI confidence score to route content automatically. High-confidence jobs skip manual checks, while Linguists only review the low-quality segments.

Enable Intelligent Workflow

Note

Currently, it is only possible to make the Intelligent Workflow settings during project creation. You cannot change them in the Project Editor.

This setting is only displayed if the Intelligent Workflow feature has been activated for your instance by your XTM Cloud Administrator. For details, see Global Intelligent Workflow settings.

These conditions must be met so that you can use Intelligent Workflow in the project:

  • the Use AI Translate & TQI checkbox has been selected. Otherwise, the Enable Intelligent Workflow setting is displayed, but inactive.

  • The After analysis option must be selected for the Send text for translation and scoring setting in the AI Translate & TQI settings. For details, see Managing AI Translate & TQI at global level.

  • In the project creation form, the Populate target with match setting in the Machine translation section is set to either Always, fixing inline differences or Only if inlines are the same. For details, see the "Populate target with match" table row below.

The Intelligent Workflow feature leverages AI Translate & TQI to power your workflow automation. It helps you save time and resources by closing unnecessary workflow steps automatically in jobs whose average TQI exceeds the custom Acceptable threshold. A job's average TQI is calculated by adding the scores from all segments and dividing them by the number of segments that have received a score. ICE matches, leveraged (100%) matches, and manual translations made by a Linguist are not evaluated by the AI Translate & TQI feature. For this reason, they are not taken into account when calculating a job's average TQI.

However, these two additional requirements must also be met for a workflow step to be closed automatically:

  • All target segments must be populated with content, either from a translation memory (TM) match or by being translated automatically by AI Translate & TQI. If any target segment in a file is empty, workflow steps cannot be closed automatically for that file.

  • All target segments in a file must be in the Good quality range. If any target segment in a file has been scored as Non-Acceptable or even Acceptable, workflow steps cannot be closed automatically for that file.

Example 26. Calculating a job's average TQI

A project's TQI Acceptable threshold is set to 70%. This means segments whose TQI is between 70% and 100% are considered to be in the Good quality range. The source file for a job contains six segments. The first three segments are translated by LLM, with TQI results of 90%, 65% and 70%. Segment 4 has been translated manually by a Linguist. Segment 5 has been filled automatically with an ICE match. Segment 6 was initially LLM-translated, but a Linguist then modified it, so it no longer has a TQI. In this case, the job's average TQI is calculated solely on the basis of the first three segments, as follows:

(90% + 65% + 70%) : 3 = 225 : 3 = 75%

The job's average TQI exceeds the project's Acceptable threshold. However, one of the segments has a TQI result of 65% which is below the Acceptable threshold. This means that workflow steps selected for the Intelligent Workflow feature will not be closed automatically. In situations like this, we would recommend that a Linguist should review the TQI suggestions about this lower-quality segment, in XTM Workbench, and correct the segment accordingly.


Important

For the best user experience, we recommend using the Intelligent Workflow feature with the new Workflow Editor (Beta). For details, see Project Editor – Beta Workflow Editor.

Optional

Checkbox

Valid values:

  • cleared – the Intelligent Workflow feature is deactivated for the project.

  • selected – the Intelligent Workflow feature is activated for the project. When you select the checkbox, the Automatically close steps setting is displayed.

Automatically close steps

Note

Currently, it is only possible to make the Intelligent Workflow settings during project creation. You cannot change them in the Project Editor.

This setting is only displayed if you have selected the Enable Intelligent Workflow checkbox.

Select which workflow steps are to be closed automatically if a job's average TQI exceeds the custom Acceptable threshold. This means that, on average, segments in the job are of Good quality.

Optional

Multiple selection

A predefined list of all steps from the workflow definitions selected in the Workflow section of the project creation form for these settings (if applicable): Main workflow, New source workflow, every Language specific workflow, and Workflow for non-analysable files. For details, see Creating a project: Workflow.

Some steps can be selected by default, as specified in system settings, but you can now deselect them and select other workflow steps, if required.

General settings for machine and LLM-generated translations

Do not search for an MT/LLM match when one of the selected matches exists

This setting is displayed at the bottom of the Machine translation section and applies to all MT engines/LLM features that have been enabled in the project.

When any of the options are selected, the segments concerned are not translated using MT/LLM.

Optional

Checkbox

Valid values:

  • ICE.

  • Leveraged.

  • Repetition.

  • 95-99% fuzzy.

  • 85-94% fuzzy.

  • 75-84% fuzzy.

  • 95-99% fuzzy repetition.

  • 85-94% fuzzy repetition.

  • 75-84% fuzzy repetition.

The default is specified in system settings or customer settings.

Note

The customer settings take priority over system settings.

Populate target with match

This setting is displayed at the bottom of the Machine translation section and applies to all MT engines/LLM features that have been enabled in the project.

Specify whether matches should be inserted in the target automatically, according to inline tag differences, and, if so, when.

Optional

Dropdown list

Valid values:

  • Never.

  • Only if inlines are the same.

  • Always, fixing inline differences.

The default is specified in system settings or customer settings.

Note

The customer settings take priority over system settings.

Set MT/LLM matches to Finished

This setting is displayed at the bottom of the Machine translation section and applies to all MT engines/LLM features that have been enabled in the project.

Optional

Dropdown list

Valid values:

  • Never (default).

  • All steps.

  • First step only.

  • All except last step.

The default is specified in system settings or customer settings.

Note

The customer settings take priority over system settings.