Hey all,

So one of the most pointless and infuriating parts of my job involves changing large portions of text from second person to third person. It’s not enough to simply use Find and Replace in Word - the converted text has to actually make sense and not be super repetitive. It seems so easy, but I can’t find an AI that can do this for large chunks of text; generally I’d need to convert around 5000 words each time, sometimes more. Because of the volume of my other work, I just don’t have time to chop it up into pieces for ChatGPT.

Does anybody know of a tool that can do this? Thanks in advance!

Edit: Supposedly Google Bard can handle up to 20,000 words, but because my documents have bullet points and other formatting it completely throws it off. Even after pasting into Notepad and back into Bard, Bard still couldn’t handle as many words as ChatGPT. It’s so frustrating, because it seems like such an easy ask.

  • mysoulishome@lemmy.world
    11·
    1 year ago

    If this your job and your employer turns a profit can’t they pay for a version of chatgpt that doesn’t have the limits?

    • CrapConnoisseur@lemmy.worldOP
      4·
      1 year ago

      I work for a university. And this is honestly something that we shouldn’t be asked for, as the group that asks for this could still understand everything perfectly well in second person. Since my job requires a lot of specialized knowledge and skills, it’s a completely stupid use of our time, but try telling them that…

  • mkhoury@lemmy.ca
    2·
    1 year ago

    Anthropic’s Claude has a really large context window. That might work for you.

    • CrapConnoisseur@lemmy.worldOP
      2·
      1 year ago

      THANK YOU! This seemed to be the best for me, with a couple of caveats (as of July 20, 2023, for anybody who has this issue in the future):

      1. When I just asked Claude if it could change a chunk of text from second to third (without giving it the text yet) it told me it couldn’t as it doesn’t have the capability to edit documents. Given its ability to process large chunks of text, that seemed strange to me, so next I just rolled it into one prompt. I told it I need a large chunk of text changed from second to third person, and then just pasted the text into the same prompt. It immediately started spitting out the text I needed!

      2. For whatever reason, it couldn’t do the full amount of text (which is around 4,900 words). However, it did about half, which is still way better than ChatGPT. So rather than stitching together something like 6 chunks of text, I only had to paste two chunks. Super simple.

      3. The output wasn’t quite as sophisticated as ChatGPT - it removed formatting, and was more repetitive in how it handled the conversion (but still far better than using Word’s Find and Replace feature, lol). For my purposes this isn’t an issue, but for others it may be.

      So thank you again for this solution. This is what I’ll be using going forward.

      • mkhoury@lemmy.ca
        1·
        1 year ago

        That’s great to hear! I don’t know if you have access to Claude 2. If you don’t then the quality of the conversion might get better soon :)

  • Z4rK@lemmy.world
    7·
    1 year ago

    You can ask ChatGPT to write you a Python program you can run that will ask ChatGPT (OpeanAi) API to perform this task for you, in chunks that are under the max limit each, and concatenate the transformed chunks to a text output file for you.

    • CrapConnoisseur@lemmy.worldOP
      1·
      1 year ago

      Thanks! For now I’m using Claude, as suggested by @mkhoury@lemmy.ca. But as soon as I get a moment I’ll try this out, as it seems like it would be the best long-term solution.

    • CrapConnoisseur@lemmy.worldOP
      2·
      1 year ago

      Thanks! I did just try it, but ChatGPT was completely arbitrary about how much it would let me feed into it. So sometimes it wouldn’t process the entire chunk that the splitter thought it should be able to process. No fault of the Splitter developers, of course - ChatGPT was just completely inconsistent about its limits. :/