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

    Phrase Next GenMT

    Content is machine translated from English by Phrase Language AI.

    Phrase Next GenMT is a Phrase-developed generative AI translation engine that leverages OpenAI’s newest large language models for automated translation. Additionally, it supports MT glossary use, tag handling and dynamic real-time customization through retrieval-augmented generation (RAG).

    Phrase Next GenMT is available as a native MT engine via Phrase Language AI and consumes MTUs. Customers do not need their own OpenAI credentials in order to start using it.

    Building on state-of-the-art generative AI technology, Phrase Next GenMT optimizes for fluency and handles terminology with ease, even in cases where an MT glossary is not attached as a reference.

    Main features

    • Generative AI Technology

      Unlike traditional neural machine translation (NMT) solutions, Phrase Next GenMT leverages generative AI technology to bring customers increased fluency and accuracy in translation, across a wider selection of domains.

    • Multi-Segment Translation

      Phrase Next GenMT processes entire text blocks instead of individual segments in a single request, resulting in more accurate and coherent translations. This eliminates errors caused by segment-level processing, such as inconsistent tone, incorrect pronouns, and terminology mismatches.

      By preserving context, multi-segment translation can improve translation quality and reduce post-editing effort.

    • Advanced MT Glossary Support

      If MT glossary is attached, instead of blind search and replace substitution, Phrase Next GenMT’s implementation supports morphological inflection, ensuring precise terminology usage.

    • Automated Tag Placement

      Supports both formatting and placeholder tags, allowing Phrase Next GenMT to work with complex sequences seamlessly.

    • Locale Variants

      Phrase Next GenMT translates fluently between locale variants such as British English (en-GB), Canadian French (fr-CA), Brazilian Portuguese (pt-BR) and Mexican Spanish (es-MX), and more.

    • Formality Control

      Phrase Next GenMT allows users to define the level of formality for the translated output. The following options can be configured for the engine in the relevant MT profile:

      • Formal

      • Informal

      • Preserve (default setting): Matches the formality of the source text.

      Formality affects vocabulary and pronouns, but its impact extends beyond grammar. It determines the overall register of the translated content: whether it reads as professional or familiar.

      Examples:

      • English to German

        The neutral English source "How can I help you?" translates differently depending on the formality setting:

        • Formal: "Wie kann ich Ihnen behilflich sein?"

        • Informal: "Wie kann ich dir helfen?"

        The pronoun shifts from Ihnen to dir, and the verb construction changes accordingly. Behilflich sein ("to be of assistance") is a formal construction that sounds stilted in casual contexts, where helfen ("to help") is the natural choice. The English source is completely neutral, yet the two German translations belong in two distinct social registers.

      • English to Japanese

        For Japanese, the formality setting switches between a polite and plain tone:

        • Plain tone (常体, jōtai): sentences end with da or dearu. Used in casual conversation, but also in formal written contexts such as academic papers and news articles.

        • Polite tone (敬体, keitai): sentences end with desu or masu. Used in business emails and in conversations with superiors, colleagues, or people you are not close to.

        The neutral English source "The meeting starts at 10." translates as:

        • Plain (常体, jōtai): "会議は10時に始まる。" (Kaigi wa jūji ni hajimaru.)

        • Polite (敬体, keitai): "会議は10時に始まります。" (Kaigi wa jūji ni hajimarimasu.)

        The meaning is identical; only the verb ending changes. Hajimaru is the plain form; hajimarimasu adds the masu ending that signals respect for the reader.

      Available languages

      Formality settings are available for all languages supported by Phrase Next GenMT.

      Note

      The formality setting is applied to segments that are freshly translated by Phrase Next GenMT. However, 1:1 matches from a translation memory (TM) will bypass the formality setting. This can lead to inconsistencies in formality within a job if the TM contains translations with a different tone.

    Real-Time Customization via RAG

    Enables real-time customization of translations using translation memories (TMs). Few-shot prompting is used with relevant TM matches to immediately adapt model behavior and ensure consistency with brand voice, style and terminology.

    How RAG works:

    1. RAG identifies and retrieves a set of fuzzy matches from the TM assigned to the project.

    2. The retrieved TM segments are used as examples (shots) for the model. Through few-shot prompting, these examples guide the translation output to ensure alignment with customer-specific requirements.

    Multi-Segment Translation

    Unlike other MT engines, Phrase Next GenMT translates with context. The engine processes multiple segments together, which enables it to translate text consistently. Multi-segment, context-aware translation reduces errors in pronouns, verb forms and formality level.

    Use Cases

    • Localization managers seeking high-quality translations without the need for model fine-tuning.

    • Dynamic workflows where brand voice, style, and terminology evolve frequently.

    • Content requiring fast customization for varied domains and audiences.

    • Localizing or adapting content from one locale variant (e.g. en-US, pt-BR) to another (e.g. en-GB, pt-PT).

    Language support

    Given the massively multilingual nature of Large Language Models, Phrase Next GenMT is a multilingual solution that addresses multiple use cases and language needs, capable of translating a wide range of languages. Although the solution supports many language combinations, we have determined a set of recommended language pairs selected to adhere to Phrase’s high-quality standards (in both translation directions):

    • English - Czech

    • English - Spanish

    • English - French

    • English - Russian

    • English - German

    • English - Italian

    • English - Dutch

    • English - Chinese Simplified

    • English - Swedish

    • English - Japanese

    • English - Portuguese

    • English - Korean

    • English - Polish

    • English - Danish

    • English - Greek

    • English - Norwegian (Bokmål)

    • English - Indonesian

    • English - Hungarian

    • English - Arabic

    • English - Turkish

    • English - Thai

    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