Under Community Review

Please use chatgpt to make trados studio more efficient and intelligent.

Please use chatgpt to make trados studio more efficient and intelligent.

1, Chatgpt can provide good translation quality much better than the current machine translation such as “language weaver” or “google”, but the current translation model for chatgpt in trados studio is still based on sentence-by-sentence model.

Could you let chatgpt to view and translation the whole and complete source text first, which can produce much better and nature translation result, and then automatically divide to target language sentence by sentence to fill in the right segments of trados studio?

 

2, The current align douments function of trados studio is complex and ineffecient. Chatgpt can understand both source language and target language. Could you use chatgpt to do the alignment for trados studio, which should be much more efficient and save a lot of time for the translator to adjust.

 

3, Sometimes, the translation memory does not include tags and various format including font, font style, font size, font color, various symbols, whether it is bold, whether the font is italicized, text highlighting etc. The translation memory is pure text. After pre-translate with this kind of translation memory with pure text, there will be a lot of work to add and adjust all the tags and various format in the target language segments.

Could you let chatgpt to do this boring and time-cosuming work for translators, since chatgpt can understand all the words and phrases from both source language and target language and chatgpt can add right tags or format in the right place.  

4, Could you make chatgpt to extract terms from sdlxliff document in trados studio? Beside during pre-translation process with machine translatioin or chatgpt, could you let chatgpt batch replace all the terms with correct formin the targer language segements? chatgpt can automatically use the right grammatic form for every terms in the target language.  

jack zhong
Parents

  •   

    This is too much of a mixed bunch to get my vote. I'd scrap this one and divide it into actionable parts.

    #1: You are basically suggesting to pre-translate the text as a whole, then do the segmentation step for both languages, then align (see #2).

    #2: Yes. This tool could profit greatly from AI integration. There is a tool in the AppStore that claims to do this, though.

    #1 and #2 would be one idea/feature request, centring around upgrading the alignment feature with AI.

    #3: This is already possible with the current AI integration, you just have to prompt it and run AI as a batch process.

    This does not need any idea or feature request. It does need some documentation though, maybe a Multifarious Filkin blog post.

    #4: I which way would this be superior to the old-fashioned, statistical way to term extraction?

    Two things rolled into one: Terminology/term replacement. This seems to overlap heavily with #3 and is possible already. Term extraction with AI: What exact functionality would that be? What more would it be than pasting the text into ChatGPT and prompting it to respond with the most important terms? How would a resulting list be integrated into the translation workflow? I find this is too vague, too much a generic “make it better with AI”. As of now, Studio does not have a built-in term extraction at all, there used to be a stand-alone application called “Term Extract”, but I think that's not maintained any more. If such a functionality is not tightly integrated, it begs the question of why it should be added in the first place. If it is to be integrated then the question is how.

    I don't want to be dismissive, but this is how I think about your proposal(s).

  • in reply to Daniel Hug

    Dear Mr. Daniel Hug

    Thank you for your reply and helping improve trados studio.

    Please allow me to explain my proposal based on some of your notes.

    #1: You are basically suggesting to pre-translate the text as a whole, then do the segmentation step for both languages, then align (see #2).

    Remark: The current mode of trados studio is useful in many applications, especially with the ability to produce an accurate translation memory with sentence by sentence. However, after understanding the whole text and then translate, it is the natural and much better way for even machine translation, chatgpt can achieve this quickly. the traditional mode of translating sentence by sentence is ok, but the ability to translate the whole passage with chatgpt would be much better since chatgpt understand the whole context and background.

    My suggestion is to add chatgpt as a machine translation plugin, trados stuido can not only allow chatgpt to translate sentence by sentence, but also let chatgpt to translate the whole passage first and then automatically divide or align all the target segments in the “edit” interface of trados studio quickly.

     

    #3: This is already possible with the current AI integration, you just have to prompt it and run AI as a batch process.

    Remark: Could you let know the name of the tools that can quickly add all the tags and automatically adjust formats in the target segments. In addition, trados studio might be able add some tags but maybe more than 1000 translation units are required, which is not practical in many small translation projects.

     

    #4: I which way would this be superior to the old-fashioned, statistical way to term extraction?

    Remarks: There are some tools which can extract terms from word documents on about term extraction, but only a small proportion of the extracted terms are useful. Moreover, we hope the translators can have the turn-key or one-stop solution in trados studio.

    What’s more, at present, when we use machine translation with term replacement, we might meet some problems. for example, in the terminology, source language “红儿“, and target language : “red flower”. The “red flower” in the target language could appear as: “ red flower” “red flowers”, “Red Flower”, “the red flower”, “the red flowers”, and so on. translators can continue to spend a lot of time  to check and correct the term in the target language one by one, or, if possible, why not invite chatgpt in the trados studio to complete the times-consuming work in a few seconds.

  • in reply to jack zhong

    automatically add all the tags in the target segments

    Screenshot of Trados Studio showing a segment with a hyperlink in the source and target text. The target text includes transliterated Chinese characters.

    automatically adjust formats in the target segments

    Screenshot of Trados Studio with a warning message in the target segment instructing to keep original fonts, styles, font sizes, and other formatting in the translated text.

  • in reply to jack zhong

     

    I am aware that the combination English-German might be very different from English-Chinese, but from my experience, the tags are copied fine when I machine-translate a segment. Fonts and styles are tags just like hyperlinks, so there is not really a difference.

    There is a plugin “AI Assistant”, which you can instruct to add tags. You can't force it to leave the translation untouched, that is still a shortcoming. To have a built-in “auto-tag” functionality would be nifty. (There is one that came with Studio 2021, but it does not work well for me.)

    AI-powered auto-tagger would be a good idea.

  • in reply to jack zhong

     

    Concerning #1: I think I understand what you mean. It's the same in German, you may have a reference to a previous segment that can only be translated correctly in the context of that previous segment. I support this idea, but I would parcel it out as a separate idea for people to vote on.

    Concerning #3: As in my post below, in most cases MT does all the tag insertion correctly. Using the AI Assistant plugin you can instruct Studio to insert tags, but there is a caution that it does not yet know how to leave the target text untouched. I hope this will come. Upgrading of the current auto-tagger function with AI would be a good idea to post individually. Using the AI Assistant I found that it does not necessarily understand where additional spaces need to be between words etc. Your use-case is insertion from TMs with different, complex styling. I will have an eye on this and see how well my prompt works.

    Concerning #4: might be a problem that is not prevalent in my language pair(s). Words entered into MT glossaries are correctly inflected when applied – actually more so than the terminology correction function of the AI Assistant. Obviously, German and English are related languages, which might make things easier.

Comment
  • in reply to jack zhong

     

    Concerning #1: I think I understand what you mean. It's the same in German, you may have a reference to a previous segment that can only be translated correctly in the context of that previous segment. I support this idea, but I would parcel it out as a separate idea for people to vote on.

    Concerning #3: As in my post below, in most cases MT does all the tag insertion correctly. Using the AI Assistant plugin you can instruct Studio to insert tags, but there is a caution that it does not yet know how to leave the target text untouched. I hope this will come. Upgrading of the current auto-tagger function with AI would be a good idea to post individually. Using the AI Assistant I found that it does not necessarily understand where additional spaces need to be between words etc. Your use-case is insertion from TMs with different, complex styling. I will have an eye on this and see how well my prompt works.

    Concerning #4: might be a problem that is not prevalent in my language pair(s). Words entered into MT glossaries are correctly inflected when applied – actually more so than the terminology correction function of the AI Assistant. Obviously, German and English are related languages, which might make things easier.

Children
No Data