Machine Translation
    1. Phrase
    2. Global
    3. Machine Translation

    Phrase Language AI

    Content is machine translated from English by Phrase Language AI.

    Phrase Language AI is a machine translation (MT) and agentic AI translation hub with additional features. It acts as the central MT layer across the Phrase Platform and enables automated translation workflows across multiple products.

    Machine translation can start with Phrase Language AI's native engines, or supported engines can be added via an API key. The AI-powered MT Autoselect feature finds the optimal engine for each translation job based on its domain and language pair.

    Phrase QPS supporting many language pairs is enabled for all engines by default. Glossaries can also be associated with MT engines that support them.

    Phrase Language AI also enables agentic AI workflows via AI Translation Agent, combining machine translation and post-editing in one single step. The AI Translation Agent is not configured as an MT provider in Phrase Language AI, but becomes available in projects and related settings once multiple MT profiles are enabled.

    • Multiple instances of Phrase Language AI cannot be created.

    • Formality levels can be set in some of the native engines, such as Phrase Next GenMT.

    • All native and ecosystem engines are presented on the Phrase Language AI Dashboard tab or, if MT profiles are enabled, on the Phrase Language AI profile page. Native and ecosystem engines can be switched on and off from these locations.

    • MT engines available as native engines under Phrase Language AI cannot be added to Phrase Language AI as an ecosystem engine.

    • The MT profiles tab displays information about engines, projects or glossaries included in the relevant MT profile.

    Cross-product usage

    Phrase Language AI is used across multiple Phrase products:

    • Phrase TMS

      Provides MT for enterprise localization workflows.

      MT profiles and engine configuration are managed in Phrase TMS, even when Phrase Language AI is used in other Phrase products.

    • Phrase Strings

      Provides MT for software localization workflows, with support for MT glossaries and integration into the Strings editor.

    • Phrase Portal

      Enables file-based translation workflows through a secure, customizable interface using configured MT profiles and engines.

    • Phrase Studio

      Supports media localization workflows leveraging MT profiles and translation resources.

    • Phrase Language AI via API

      Allows the deployment of the Phrase Language AI machine translation solution outside the Phrase Platform, enabling high-quality MT and new localization solutions at scale. It supports:

      The minimum requirement for using Phrase Language AI via API is the Phrase Portal user role.

    MT Autoselect Supported Languages

    MT Autoselect supports domain-based recommendation for all documents in these source languages:

    • English

    • Czech

    • German

    • Spanish

    • French

    • Italian

    • Japanese

    • Korean

    • Dutch

    • Russian

    • Swedish

    • Chinese Simplified

    All other language combinations get MT Autoselect recommendations but do not take text domain into account.

    Domains

    The type of content translated with machine translation can significantly affect the quality of the output. An engine that might perform well for product descriptions might struggle with a medical report. The content type variable is known as a domain.

    Categorizing documents into domains is normally a process that requires human input, creating complications when it comes to scaling translation and allows for basic human error.

    Domains were created automatically using an AI-powered analysis of documents. An unsupervised machine learning algorithm recognized 11 distinct types of documents that shared similar sets of keywords:

    Domain

    Keywords

    Medical

    ‘study’, ‘patients’, ‘patient’, ‘treatment’, ‘dose’, ‘mg’, ‘clinical’

    Travel and Hospitality

    ‘km’, ‘hotel’, ‘guests’, ‘room’, ‘accommodation’

    Business and Education

    ‘team’, ‘business’, ‘work’, ‘school’, ‘students’

    Legal and Finance

    ‘agreement’, ‘company’, ‘contract’, ‘services’, ‘financial’

    Software User Documentation

    ‘click’, ‘select’, ‘data’, ‘text’, ‘view’, ‘file’

    Consumer Electronics

    ‘power’, ‘battery’, ‘switch’, ‘sensor’, ‘usb’

    User Support

    ‘please’, ‘email’, ‘account’, ‘domain’, ‘contact’

    Cloud Services

    ‘network’, ‘server’, ‘database’, ‘sql’, ‘data’

    Industrial

    ‘mm’, ‘pressure’, ‘valve’, ‘machine’, ‘oil’

    Software Development

    ‘value’, ‘class’, ‘type’, ‘element’, ‘string’

    Entertainment

    ‘game’, ‘like’, ‘get’, ‘love’, ‘play’, ‘go’

    Configure Phrase Language AI

    Important

    MT profiles and engine configuration are managed in Phrase TMS, even when Phrase Language AI is used in other Phrase products.

    Native and ecosystem machine translation engines can be activated and deactivated from the Phrase Language AI page (access from the left-hand navigation panel in Phrase TMS) by TMS Administrators and Project Managers with required permissions. If MT profiles are enabled, AI Translation Agent can also be added as MT provider in projects, templates or Phrase Portal.

    Additional engines can be added if the API key for the engine is available.

    To add an additional engine, follow these steps:

    1. From the Phrase Language AI page, click Connect MT engines.

      The Connect MT engine page is displayed.

    2. Select an engine.

      The page for the selected engine opens.

    3. Provide the API Key.

    4. If required, provide additional information for the engine.

    5. Click Save.

      The configured engine is available for activation.

    BYO Engine

    The Bring Your Own Engine (BYO Engine) feature allows the integration of external machine translation (MT) engines into Phrase to be used in translation workflows like any other MT provider.

    Custom-trained engines can be used or translations can be routed to in-house LLMs or MT providers to maintain full control over quality, privacy, and cost while continuing to benefit from Phrase’s translation memories (TM), glossaries, QPS and the CAT editor.

    This is ideal for:

    • Enterprises with domain-specific MT models.

    • LSPs with custom client engines or preferred AI stacks.

    • Organizations with internal policies that require using specific engines (e.g. Gemini, internal LLMs, etc.).

    Requirements

    • The client (or organization) needs to build and test a lightweight API adapter. This adapter enables communication between Phrase and the engine.

      The adapter must support:

      • Translation requests

      • Language pair support

      • Engine status check (Phrase sends a status request)

    • The backing infrastructure (MT engine or LLM) must be hosted and maintained by the client/organization.

    • Phrase needs permissions to connect to that infrastructure.

    • Translations must be returned with the same count and segment order.

    Once the API adapter is implemented and deployed, the external custom MT engine becomes available in Phrase Language AI for selection in MT profiles and can be assigned to projects.

    FAQ

    • Why is the status request required? Is it sent by Phrase to the engine, or is it expected to be handled by the customer?

      Phrase sends the request to check whether the engine works or not.

    • One request may contain 500 maxItems in segments. Does this imply that 500 source sentences must be translated in real time?

      Synchronous calls include only up to 5 segments, while asynchronous calls can include more, up to 500. Also, the async timeout is about 30 minutes.

    • Are all terms of a glossary sent in a request, or a subset?

      All terms of the glossary are sent.

    • Is it possible to plug in an LLM?

      It should be possible with a wrapper. As long as the engine returns segments as defined in the schema, Phrase does not require a specific engine or LLM.

    • Is custom quality evaluation (QE) possible?

      Third-party QE is not supported.

    • How are tags handled?

      The content Phrase receives is pre-processed. HTML tags are converted into marks, and those marks are passed through as is.

    • Can we implement only the async approach? Will it work correctly in all Phrase areas (pre-translation, Editor, etc.)?

      No, both approaches (ideally all endpoints) must be implemented. The synchronous endpoint is used, for example, by the Editor, while the asynchronous one is used in areas such as pre-translation.

    • What happens when the engine gets multiple requests at the same time?

      The engine must be able to handle this. Ensure proper handling of concurrency and thread safety across all endpoints, particularly asynchronous ones.

    • How does Request Metadata work?

      It must remain static, as defined during engine setup.

    • How does Segment Metadata work?

      Currently, it needs to be part of the schema.

    • During MT processing, does BYOE use RAG to retrieve context?

      No, BYOE does not use RAG. Users can leverage TMs to determine which segments are translated, but they are not used within the MT process itself. This is because MT engines must be specifically designed to incorporate TM data via RAG, and in BYOE scenarios there is no control over how engines are built or configured.

    MT Profiles

    Available for

    • Team, Professional, Business, Enterprise and Software UI/UX plans

    Get in touch with Sales for licensing questions.

    Important

    DeepL will no longer be available as a native engine from Jun 30, 2026. Enable Phrase Next GenMT or continue using DeepL via Phrase Ecosystem. If no action is taken, Phrase Next GenMT will be enabled automatically. Learn more.

    MT profiles can be created with different settings and used across Phrase products, including TMS, Strings, Portal, Studio, and Phrase Language AI via API.

    If multiple MT engines are enabled in a profile, Autoselect selects the most suitable one for the job.

    If only one specific engine is to be used for MT, either:

    • Disable all others engines, or

    • Create a dedicated MT profile with only the required engine and select that profile for machine translation across Phrase products.

    To create a machine translation profile, follow these steps:

    1. From the Machine translation engines panel on the Phrase Language AI page, click Enable multiple profiles.

      The Machine translation profiles table is presented.

    2. Click Add MT profile.

      The New Phrase Language AI profile window opens.

    3. Provide a name for the profile and click Create.

      The configuration page for the profile is presented.

    4. Configure profile as required and click Phrase Language AI at the top of the page.

      The new profile is presented in the MT profiles tab.

    Profiles can be edited, assigned as default, deleted or duplicated from the More ellipses.png menu to the right of the profile name.

    If using Phrase Language AI via API, the MT profile ID is required for translation requests and can be found in the URL of the specific MT profile page.

    Example URL: https://cloud.memsource.com/tms/language-ai/profile/123abc456xyz

    123abc456xyz is the MT profile ID.

    Non-Translatable Tags

    Non-translatable tags enable parts of a text to be excluded from machine translation (MT) when processed through Phrase Language AI in TMS, Strings and Phrase Portal.

    This functionality helps preserve brand names, technical terms, and other terminology that must remain in the source language.

    PLAI supports non-translatable tags for the following native MT engines:

    • Amazon Translate

    • DeepL (not supported in Phrase Portal)

    • Google Translate

      No native API support for non-translatable tags. Results may vary.

    • Microsoft Translator

    • Rozetta T-400 Realtime

    • Widn.AI

    Non-translatable tags are set up in different ways when using PLAI in Phrase TMS, Strings and Phrase Portal:

    • Phrase TMS

      TMS administrators can configure tags for different file types at the organization level through the Convert to Phrase TMS tags option in file import settings.

    • Phrase Strings

      To exclude text from translation when uploading keys, wrap it with the <notranslate> tag, as in the example:

      I would like to <notranslate>slide<notranslate> in a water park.

    • Phrase Portal

      Tags configured in TMS are automatically reused.

    During processing, PLAI automatically converts these tags into a format recognized by the selected MT engine and reverts them back after translation to restore the original text.

    Limitations

    • Not supported for .TXT files or free text entered in the Phrase Portal interface.

    Supported Native Machine Translation Engines

    • Amazon Translate, also supported by Phrase Language AI via API.

    • DeepL and Deepl Next Gen, also supported by Phrase Language AI via API.

      Not supported in Phrase Portal.

      Important

      DeepL will no longer be available as a native engine from Jun 30, 2026. Enable Phrase Next GenMT or continue using DeepL via Phrase Ecosystem. If no action is taken, Phrase Next GenMT will be enabled automatically. Learn more.

    • Google Translate, also supported by Phrase Language AI via API.

    • Microsoft Translator / Microsoft Translator Hub, also supported by Phrase Language AI via API.

    • Phrase NextMT, also supported by Phrase Language AI via API.

    • Phrase Next GenMT, also supported by Phrase Language AI via API.

    • Rozetta T-4OO Realtime, also supported by Phrase Language AI via API.

    • Tencent TranSmart, also supported by Phrase Language AI via API.

    • Widn.AI, also supported by Phrase Language AI via API.

    Supported Ecosystem Machine Translation Engines

    • Alexa Translations A.I. (formerly Yappn)

    • Custom NextMT

    • DeepL

    • Google AutoML, also supported by Phrase Language AI via API.

      Note

      When adding Google AutoML to Phrase Language AI, Bucket name is required information due to support and potential usage of MT glossaries.

    • Human Science

    • Hunyuan

    • Lengoo HALOS

    • Microsoft Custom Translator (in Microsoft Azure, set location to "global"), also supported by Phrase Language AI via API.

    • ModernMT

      Note

      A ModernMT license that allows usage within CAT tools is required. To retrieve information about the license, use the /users/me endpoint.

    • NpatMT

    • Omniscien MT V6

    • Systran PNMT

    • Ubiqus NMT

    • Yandex

    Was this article helpful?
    ★ ★ ★ ★ ★

    Sorry about that! In what way was it not helpful?

    The article didn’t address my problem.
    I couldn’t understand the article.
    The feature doesn’t do what I need.
    Other reason.

    Note that feedback is provided anonymously so we aren't able to reply to questions.
    If you'd like to ask a question, submit a request to our Support team.
    Thank you for your feedback.

    Articles related to Machine Translation