OpenAI Translator

Dear Trados community,

I'm currently looking into the OpenAI Translator App and ways to create new prompts in addition to the default ones that may be useful to our pool of translators.
For example, I created one prompt that proofreads the translator's own version and suggests two additional alternatives:

"This text was translated from {0} into {1} by a native professional translator: {2}. Please provide a proofread version of that translation as well as two paraphrased versions."

This seems to be working fine as a basic first step (apart from the gpt4 engine not despite having a Plus account), but taking a closer look at the definition of the placeable parameters {0}, {1} and {2} I'm not sure I've done this correctly/ am taking the right approach:

I can choose the search text {2} to be either source text or translation and the language {0} will adapt automatically, I assume. But either way placeable {1} will always be the target language.


Since my prompt uses the translation as search text, am I basically saying "This text was translated from {Italian} into {Italian}?" Maybe I could just leave out the first sentence and be done with it, but I'd very much like to receive a proofread version and two additional suggestions (that takes into account the source text) with only one prompt, but for that, my prompt should be able to say "from English to Italian". I don't think this works as of now, as you can only define one search text at the moment.

In case I've caused more confusion than clarity, I'll try and break this down into two "needs" I think are required for this to work and which might exponentially increase the possibilities of the app:

1. To have additional placeable parameters for source and target language that do not change based on the selected search text.
2. To have more than one search text placeable parameter, e.g. one for proofreading the translation in itself, and one for looking at the source text and suggesting two translations that are different from the one already there.

Or maybe I'm just overlooking something basic, so kindly let me know. Looking forward to your suggestions!

Kind regards,
Guillaume

emoji
Parents
  •  

    Yes, I think you're doing this correct, and also think the app could be expanded to have more options ( fyi).  You should make the suggestion in the ideas site:

     RWS AppStore Ideas 

    In the meantime I think you have two ways to achieve your aim... both a bit messy and one you already did.  You probably tried them both!

    Proofread and suggest corrections for this translation from {0} to {1}. Also, provide two alternative translations.: {2}. 

    Which would deliver something like this:

    OpenAI UI showing the results from the prompt.

    Or this:

    Proofread and suggest corrections for this translation from {0} to {1}: {2}. Also, provide two alternative translations.

    Which would get you this:

    Screenshot showing the results of the next prompt.

    The second one is essentially the approach you have taken and is better albeit with a redundant part:

    Screenshot showing a  redundant output that misuses a translation by adding the title of the section.

    Sort of workable, but for your requirement (which seems reasonable), the UI could definitely be enhanced.  A nice usecase.

    Paul Filkin | RWS Group

    ________________________
    Design your own training!

    You've done the courses and still need to go a little further, or still not clear? 
    Tell us what you need in our Community Solutions Hub

    emoji
Reply
  •  

    Yes, I think you're doing this correct, and also think the app could be expanded to have more options ( fyi).  You should make the suggestion in the ideas site:

     RWS AppStore Ideas 

    In the meantime I think you have two ways to achieve your aim... both a bit messy and one you already did.  You probably tried them both!

    Proofread and suggest corrections for this translation from {0} to {1}. Also, provide two alternative translations.: {2}. 

    Which would deliver something like this:

    OpenAI UI showing the results from the prompt.

    Or this:

    Proofread and suggest corrections for this translation from {0} to {1}: {2}. Also, provide two alternative translations.

    Which would get you this:

    Screenshot showing the results of the next prompt.

    The second one is essentially the approach you have taken and is better albeit with a redundant part:

    Screenshot showing a  redundant output that misuses a translation by adding the title of the section.

    Sort of workable, but for your requirement (which seems reasonable), the UI could definitely be enhanced.  A nice usecase.

    Paul Filkin | RWS Group

    ________________________
    Design your own training!

    You've done the courses and still need to go a little further, or still not clear? 
    Tell us what you need in our Community Solutions Hub

    emoji
Children
  • Hi Paul, thank you for your lightning-swift reply. I'll try out your prompt versions and see what it gets me/which one I prefer.

    My version also gives me these redundant parts, but I actually think it structures the output quite nicely if you turn off "Show differences". Only then you won't have the tracked changes for the proofread part, but as long as we can't separate these two tasks more clearly, ticking that checkbox is only a small effort.

    I'll add my suggestions to the ideas site and try to forumlate them clearly enough.

    Thanks!

    emoji