How can I train a gpt-35-turbo model with an internal company document that is available online and periodically updated?

Question

I want to be able to create an interactive chatgpt that is able to read the contents of the specific web document links I provide to it. How do I train a chatpgt model based on contents from a web document? I only want to pass url pointing to the location where the web document is based.

From a user perspective, instead of users going and reading the document to find a specific piece of information, they can just write theiir wuestion with some key words, and Azure Open AI should be able to automatically find and show that piece of information from the document based on the url's that were provided for training.

Your support will be highly appreciated

Answer 1

Sharma, Patrick Greetings & Welcome to Microsoft Q&A forum!

I want to be able to create an interactive chatgpt that is able to read the contents of the specific web document links I provide to it. How do I train a chatpgt model based on contents from a web document? I only want to pass url pointing to the location where the web document is based. From a user perspective, instead of users going and reading the document to find a specific piece of information, they can just write theiir wuestion with some key words, and Azure Open AI should be able to automatically find and show that piece of information from the document based on the url's that were provided for training.

There are several steps involved in this scenario. I would suggest you, please go through the below information and let us know if you have any further queries.

To train a ChatGPT model based on the contents of a web document, you will need to extract the text content of the document and format it in a way that the ChatGPT model can understand. Here are the general steps you can follow:

• Extract the text content: You can use a web scraping tool to extract the text content of the web document and save it to a file or database.
• Prepare the training data: You will need to prepare the training data by formatting it in a way that the ChatGPT model can understand. This involves using the Chat Markup Language (ChatML) format, which is a specific chat-like transcript format that the ChatGPT model expects. You can use the extracted text content of the web document to create a system message that provides context and instructions to the model, followed by a series of messages between the user and the assistant.
• Train the model: You can use the Azure OpenAI Service to train the ChatGPT model on the prepared training data. You will need to provide the model with the URL of the web document and the user's query, and the model will generate a response based on the contents of the document.
• Deploy the model: Once the ChatGPT model is trained, you can deploy it to a web application or other platform to make it available to users.
• Test and refine the model: You will need to test the ChatGPT model and refine it as necessary to improve its accuracy and performance. This may involve fine-tuning the model, adjusting the training data, or making other changes to the model or the application.

Make sure that you have the necessary permissions and rights to access and use the web document you are scraping, and that you are complying with all applicable laws and regulations.

References:
Learn how to customize a model for your application
Create a resource and deploy a model using Azure OpenAI

Learn how to work with the ChatGPT and GPT-4 models

Get started generating text using Azure OpenAI Service

If this answers your query, do click Accept Answer and Yes for was this answer helpful. And, if you have any further query do let us know.