Bring your own data to Azure OpenAI chat models

by | Jun 22, 2023 | Technology | 0 comments

This article is contributed. See the original author and article here.

Introduction


Azure OpenAI models provide a secure and robust solution for tasks like creating content, summarizing information, and various other applications that involve working with human language. Now you can operate these models in the context of your own data. Try Azure OpenAI Studio today to naturally interact with your data and publish it as an app from from within the studio.


 


Getting Started


Follow this quickstart tutorial for pre-requisites and setting up your Azure OpenAI environment.


 


In order to try the capabilities of the Azure OpenAI model on private data, I am uploading an ebook to the Azure OpenAI chat model. This e-book is about “Serverless Apps: Architecture, patterns and Azure Implementation” written by Jeremy Likness and Cecil Phillip. You can download the e-book here


 


Before uploading own data


Prior to uploading this particular e-book, the model’s response to the question on serverless design patterns is depicted below. While this response is relevant, let’s examine if the model is able to pick up the e-book related content during the next iteration


 


pre-training.png


 


After uploading own data


This e-book has an exclusive section that talks in detail about different design patterns like Scheduling, CQRS, Event based processing etc.


 


ebook.png


After training the model on this PDF data, I asked a few questions and the following responses were nearly accurate. I also limited the model to only supply the information from the uploaded content. Here’s what I found.


 


post-training.png


 


Now when I asked about the contributors to this e-book, it listed everyone right.


 


post-training-1.png


 


Read more


With enterprise data ranging to large volumes in size, it is not practical to supply them in the context of a prompt to these models. Therefore, the setup leverages Azure services to create a repository of your knowledge base and utilize Azure OpenAI models to interact naturally with them.


 


The Azure OpenAI Service on your own data uses Azure Cognitive Search service in the background to rank and index your custom data and utilizes a storage account to host your content (.txt, .md, .html, .pdf, .docx, .pptx). Your data source is used to help ground the model with specific data. You can select an existing Azure Cognitive Search index, Azure Storage container, or upload local files as the source we will build the grounding data from. Your data is stored securely in your Azure subscription.


 


We also have another Enterprise GPT demo that allows you to piece all the azure building blocks yourself. An in-depth blog written by Pablo Castro chalks the detail steps here.


 


Getting started directly from Azure OpenAI studio allows you to iterate on your ideas quickly. At the time of writing this blog, the completions playground allow 23 different use cases that take advantage of different models under Azure OpenAI.


 



  1. Summarize issue resolution from conversation

  2. Summarize key points from financial report (extractive )

  3. Summarize an article (abstractive)

  4. Generate product name ideas

  5. Generate an email

  6. Generate a product description (bullet points)

  7. Generate a listicle-style blog

  8. Generate a job description

  9. Generate a quiz

  10. Classify Text

  11. Classify and detect intent

  12. Cluster into undefined categories

  13. Analyze sentiment with aspects

  14. Extract entities from text

  15. Parse unstructured data

  16. Translate text

  17. Natural Language to SQL

  18. Natural language to Python

  19. Explain a SQL query

  20. Question answering

  21. Generate insights

  22. Chain of thought reasoning

  23. Chatbot


Resources


There are different resources to get you started on Azure OpenAI. Here’s a few:



 

Brought to you by Dr. Ware, Microsoft Office 365 Silver Partner, Charleston SC.

Submit a Comment