• filcuk@lemmy.zip
    link
    fedilink
    English
    arrow-up
    3
    ·
    2 days ago

    Why would they use LLMs for translation over dedicated translator (apps like Google translate)? Is this common now?

    • daniskarma@lemmy.dbzer0.com
      link
      fedilink
      English
      arrow-up
      4
      arrow-down
      1
      ·
      edit-2
      1 day ago

      In my experience LLM get vastly better results that traditional translation software.

      Also google translate is not traditionally suited for long coherent text. One particular issue is tone, proper translation takes into account not only the worlds but the tone and the subject is being treated. Google translate cannot take that into account, with an llm the user can tweak the tone to match the tone of the original text with better accuracy.

      And, anyway google translate if it’s not already using llm for translation will soon. Results are just better. It’s one of the tasks that language models are actually good for.

      Anyhow, what’s the issue if an automatic translation is done using one software or other? Just use whatever gives best results and it’s more convenient for the developer.

      • filcuk@lemmy.zip
        link
        fedilink
        English
        arrow-up
        2
        ·
        21 hours ago

        I don’t have an issue, it was just a question.
        I don’t ever translate anything but a few words myself.
        My assumption was that a dedicated tool would have done a better job, but you have good points about tone and coherence of long text (and possibly even across many promps of the translation).

    • Agent Karyo@lemmy.worldOP
      link
      fedilink
      English
      arrow-up
      4
      ·
      2 days ago

      I work with multiple languages and I haven’t heard of LLMs being used instead of Deeply/Google translate.

      Don’t really see the reason to switch to LLMs.