Microsoft QnA Maker – A Bot-Powered Knowledge Base
What if you wanted to provide a knowledge base with an intelligent, AI-based bot to your internal or external customers? QnA maker can help you with that in a fraction of time. No need to manually set up a bot project, machine learning models, storage accounts etc.
QnA Maker is a cloud-based API service that creates a conversational, question and answer layer over your data. It enables you to use semi-structured data such as FAQ URLs, documents containing manual and support information and any custom questions and answers to create a natural language understanding KB Bot. QnA Maker offers a complete no-code experience to create a FAQ Bot. It’s based on the Microsoft Bot Framework, Azure Cognitive Services, Azure Search Service and Azure App Service.
For my demo I’m using the free tier. For more information visit https://azure.microsoft.com/en-us/pricing/details/search/
How To Get Started
First, you need an Azure Subscription. Any subscription is suitable, no matter if it is a free, Pay-as-you-go or EA subscription. Before you can create any QnA Maker knowledge bases, you must first set up a QnA Maker service in Azure. You need subscription contributor permissions, or contributor permissions on an existing resource group. You can use the Azure Portal or an ARM Template to provision the service. If you struggle with ARM Template in general, we might help you out. See our ARM Templates Service Offering for more information. By adding a QnA Maker Service, Azure Resource Manager provisions a few services in the background, the QnA Maker depends on. After the deployment succeeds you find the following resources in your RG.
Next, you have to visit the qnamaker.ai portal and login with your Azure / Tenant credentials and choose “Create a knowledge base”. I’d recommend to create the QnA Service part separately as described earlier in this post.
After you selected the correct values, you have to give the KB a name and populate some data. You can of course add and remove information / data at any point later on.
Finally, create the KB.
After Initial extraction of QnA pair data, you can enhance and delete the content.
Make sure you test the bot / KB before publishing it. If it does not behave, just train it more and more 😊
If you’re satisfied with the result, publish the KB.
The information you get back can now be used to create a custom bot.
As a first step, I’ve created a little PowerShell function to test the KB Service.
As we can see, it’s working like charm. The answer contains the plain text answer, the accuracy and the source. Fine so far, let’s finally create the bot for the knowledge base.
After the bot has been created successfully, we can test its functionality.
Using Azure AI services has never been easier. Without touching any line of code I’ve showed you how to create a QnA knowledge base, ready to use in literally 10 minutes.
Call To Action
So, what's next?
Learn more about AI on the official Microsoft AI website > https://www.microsoft.com/en-us/ai
Visit the official QnA maker website > https://www.qnamaker.ai/
Get in contact with us to schedule an appointment if you want to learn more about Azure AI > firstname.lastname@example.org
Go try it out yourself! Cheers, Michael