Classify data in Google Sheets with ChatGPT

Поделиться
HTML-код
  • Опубликовано: 3 дек 2024

Комментарии • 50

  • @arturkern9320
    @arturkern9320 Год назад +1

    Your work done here is amazing. I m currently wrapping my head about some solutions and your work was a huge part of it. Thank you for your help.

    • @gptforwork
      @gptforwork  Год назад +1

      thank you! I'd be happy to hear more about your use-cases if you want to write to me.
      also do not hesitate to leave us a nice review here: workspace.google.com/marketplace/app/gpt_for_sheets_and_docs/677318054654 as it really helps!

  • @2usa5
    @2usa5 Год назад +1

    Thank you - an incredible tool. Tons of efficiency gained.

    • @gptforwork
      @gptforwork  Год назад

      You're welcome! Don't hesitate to leave us a nice review here: workspace.google.com/marketplace/app/gpt_for_sheets_and_docs/677318054654 :)

  • @Bee_Hive_10134
    @Bee_Hive_10134 9 месяцев назад +2

    This is great. If I want it to reference two columns in my Reference tab... a category and its keywords, do I have to add an ARRAY function to this GPT_CLASSIFY function as well instead of just a single column?

    • @gptforwork
      @gptforwork  8 месяцев назад

      Hello, not sure I fully understand the question, but you can include a range of cells for categories if you want

  • @Dandifederico
    @Dandifederico Год назад +5

    Thanks for this great video Stan. If Chatgpt does not get the category correct, is it possible to "train" it on your own definition or examples of the Category?

    • @gptforwork
      @gptforwork  11 месяцев назад

      Hello! Yes it's possible, with the examples parameter of GPT_CLASSIFY, or with custom instructions that can be added in the sidebar.

  • @markmancini5238
    @markmancini5238 Год назад +2

    How can i make it give me the top 5 or top 10 categories instead of one (GPT CLassify) or the 30 its giving me (GPT Tag)?

    • @gptforwork
      @gptforwork  Год назад +2

      Hi Mark, could you let me know if the method I send you via email works? Would like to repost it here if it does.
      Cheers

  • @drmattbarney
    @drmattbarney Год назад +1

    I'm a psychologist - do you recommend using CLASSIFY if I want GPT to return a single integer that reflects a specific behaviorally-anchored rating scale that clarifies whether it to give a 1, 2, 3, 4 etc. based on specific text content in a sample (value variable)?

    • @gptforwork
      @gptforwork  Год назад

      Hi Matt, it's recommended to classify with semantic categories, that you can then match with their respective integers.

  • @MeawinOtolep
    @MeawinOtolep Год назад +1

    Hello, how can I give feedback on the classification ?
    I'm using this tool to classify my expenses.
    (I download them from several banks and payment providers, consolidate them into a single Excel, and classify them in several categories).
    For the classification part, the tool is great and beats copy and paste back and forth into GPT/Sheets !
    However, I sometimes find that the classification is not working as I hoped.
    (for example, it's never able to detect when I buy coffe and classify it accordingly).
    --> With copy and paste into ChatGPT, I'm able to train the model by giving feedback to the tool. Is it something that would be implementable ?

    • @gptforwork
      @gptforwork  11 месяцев назад

      Hello! It should work with the examples parameter in GPT_CLASSIFY, or with custom instructions. Check out our documentation about GPT_CLASSIFY parameters here: gptforwork.com/help/gpt-for-sheets/gpt-functions/all-available-functions#gpt_classify-function
      We are working on new tools that will be available soon, that will save you from typing the formula. Those new tools will make them easier to add custom instructions!

  • @forsak3n87
    @forsak3n87 Год назад +2

    Is it possible to use GPT_CLASSIFY to classify column A values into very abstract groups into column B and leverage that B has already been partially filled (~20%)?

    • @gptforwork
      @gptforwork  Год назад +1

      If I understand well, I think GPT_FILL is what you're looking for. Tell me if not: ruclips.net/video/Mb6ufl4H5go/видео.html

    • @forsak3n87
      @forsak3n87 Год назад

      @@gptforwork Thanks for the response. I think the GPT_FILL is what I'm looking for but it appears that my data set is too large. I'm getting a maximum context limit of 4096 but my dataset is about 8 times that. I've tried adjusting the maximum_token variable but that did not appear to work. Is there any way around this?

    • @gptforwork
      @gptforwork  Год назад +1

      @@forsak3n87 there is probably a better way, can you fill this form with the details? docs.google.com/forms/d/e/1FAIpQLSeNW6ynH1YfmgytrbLq2tPJouuI0qCQ_azg_W3eIPXxSzFIJw/viewform

    • @forsak3n87
      @forsak3n87 Год назад +1

      @@gptforwork filled! Thanks for all the help and great tool.

    • @gptforwork
      @gptforwork  Год назад

      @@forsak3n87 just replied to you

  • @yaneletov
    @yaneletov Год назад +1

    Thank you, How many data or sentences this method can analyze for once? What if 1000 sentences?

    • @gptforwork
      @gptforwork  Год назад

      1000 sentences should be doable if you use the text-embedding-ada-002 model
      but it's probably better to do them one by one and drag the formula.

  • @koliux1
    @koliux1 Год назад +3

    in our company, a bunch of DS tried to figure that out for 1 year. WIth this a 8 year old kid can do it in 5 mins, brilliant :)

    • @gptforwork
      @gptforwork  Год назад

      hehe thanks!

    • @Jean-r8r
      @Jean-r8r 2 месяца назад

      4:11 DEEZIGNERS​@@gptforwork

  •  Год назад

    Hi, Is it possible that to classify the sentiment it can be done on a particular entity that appears in the analyzed text? How could I do it?

    • @gptforwork
      @gptforwork  11 месяцев назад

      Hello! You can specify it in your prompt if you use the =GPT() formula instead of GPT_CLASSIFY. If you're still struggling do not hesitate to contact our support team: support.gptforwork.com/hc/en-us/requests/new

  • @mykytasoima3065
    @mykytasoima3065 Год назад

    Hey! Thanks. How do you think will it work if a have a range of 6000 categories? I’m not sure that it can choose the one 😅 I use the GPT4 in paid version.

    • @gptforwork
      @gptforwork  11 месяцев назад

      Hi! It should work with gpt-4-turbo, but it will be very expensive. It would be better to group the categories to a higher level first and then subcategorize if possible.

  • @MiaoBuy
    @MiaoBuy Год назад +2

    Hey man! Really great products, can’t believe it’s free 😂.
    May I ask a dumb question.
    I know the price for OpenAI API is like 0.002 per 1000 credits and about 800 words.
    But for these prompts, it would literally just generate like 1 word, Will OpenAI charges us for the prompt as well, or we literally get to do this for that cheaply?

    • @gptforwork
      @gptforwork  Год назад +2

      Hi there. Not a dumb question. All tokens count, prompt + response. If you only generate one word (for example in a classification use-case), then you still have to pay for every word included in the prompt.
      But to be honest, it is now very absurdly cheap with gpt-3.5-turbo.

  • @ClaudWilliams
    @ClaudWilliams Год назад +1

    What is the difference between a tag and claffify?

    • @gptforwork
      @gptforwork  Год назад

      TAG returns the top_k values that apply whereas CLASSIFY returns only one value. (TAG with top_k=1 is the same as CLASSIFY)

  • @anton-dm2oz
    @anton-dm2oz Год назад

    This is absolutely brilliant

  • @NelsonBondioli
    @NelsonBondioli Год назад

    Awesome tool, however I get a lot of "#ERROR" messages, "Exceeded maximum execution time (line 0)" even when Im sorting few lines (like 10). Is there something I can do?

    • @gptforwork
      @gptforwork  Год назад +1

      Hi Nelson. Sorry about that it is due to custom functions requiring a response in less than 30 seconds. Unfortunately since 3 days it seems OpenAI servers response times are much higher, presumably because they reallocated capacity to ChatGPT Plus for GPT-4. Hopefully they get around improving availability soon.
      As for us, we should have full solution within a month, regardless of what OpenAI does.

    • @NelsonBondioli
      @NelsonBondioli Год назад

      @@gptforwork I see! I thank you for the reply/information, I'm really happy and excited with the work you guys are doing

  • @paulinaArangomar
    @paulinaArangomar 7 месяцев назад

    Hi! Very useful, works only in english?

    • @gptforwork
      @gptforwork  7 месяцев назад

      no, it works in most languages :)

    • @gptforwork
      @gptforwork  7 месяцев назад

      It works in most languages! :)

  • @MarcoDM1571
    @MarcoDM1571 Год назад

    Hey, thanks for the lessons..... as to "category" don't trust me, trust the AI:
    "Sure! The word “category” is pronounced as /ˈkætəɡɔːri/ with the stress on the first syllable. The word “categorize” is pronounced as /ˈkætəɡəraɪz/ with the stress also on the first syllable. Is there anything else you would like to know?"

  • @AshokKumar-dk2xp
    @AshokKumar-dk2xp 3 месяца назад +1

    🎉🎉🎉🎉🎉❤❤❤❤😂

  • @majiddehbi9186
    @majiddehbi9186 Год назад

    how can we get chatgpt in googla sheet clearly

    • @gptforwork
      @gptforwork  Год назад

      See the set up guide: gptforwork.com/setup

  • @samuelelouest4857
    @samuelelouest4857 7 месяцев назад

    Mw vle tandem kreyol

    • @gptforwork
      @gptforwork  4 месяца назад

      I can't speak Creol!

  • @TanbirKhan-l1e
    @TanbirKhan-l1e 16 дней назад

    Moslim

  • @largoenjapon3377
    @largoenjapon3377 Год назад +1

    Hey Stan, I just wanted to take a moment to say thank you for all of the helpful tutorial videos you've been putting out! Your expertise and attention to detail have been invaluable in helping me improve my own skills. Keep up the awesome work! #grateful #tutorialvideos #creativityboost #lifelonglearning @Stan #JuanUsingGPT

    • @gptforwork
      @gptforwork  Год назад

      You probably spent 10x more time posting on each video than writing it haha! Thanks a lot