how can we improve accuracy of existing chatbot using Azure Open AI on the place of LUIS

Question

we currently have a chatbot that uses LUIS to respond to specific questions within an intent.

The intent is based on the responses we receive from field engineers on problem tickets raised by customers. However, we have noticed that the answers to these questions are not static and change on a monthly basis.

This is due to the fact that the field engineers' responses can differ each time they encounter the same problem. As a result, the accuracy of our chatbot has decreased, and we have had to retrain LUIS with updated answers every month.

To improve the accuracy of our chatbot, Can we use Azure Open AI and how can I implement this please explain in steps?

Answer 1

Hello @Vaibhav Lambat !

Welcome to Microsoft QnA

Yes, you can use Azure Cognitive Services, specifically the QnA Maker service, to improve the accuracy of your chatbot.

The QnA Maker service allows you to create a knowledge base of frequently asked questions and their corresponding answers, and then use natural language processing to match user queries to the appropriate answer in the knowledge base. This way, you can avoid the need to constantly update your LUIS model with new responses from field engineers.

Here are the steps to implement this:

Create a QnA Maker resource in the Azure portal. This will provision the necessary resources and allow you to create a knowledge base.

Create a new knowledge base in the QnA Maker portal and add your frequently asked questions and their corresponding answers. You can import existing content from sources like an FAQ page, a document, or a CSV file, or create and edit the content directly in the QnA Maker portal.

Train the knowledge base by adding questions and their corresponding answers, and testing the knowledge base to ensure that it provides accurate answers.

Integrate the QnA Maker knowledge base with your chatbot. Depending on your chatbot platform, there may be different ways to do this. For example, if you are using the Microsoft Bot Framework, you can use the QnAMakerDialog class to query the knowledge base and return the appropriate answer.

Test the chatbot to ensure that it provides accurate answers to user queries. You may need to continue to refine the knowledge base over time as new questions and answers arise.

By using QnA Maker, you can improve the accuracy of your chatbot and reduce the need to constantly update your LUIS model with new responses from field engineers.

Kindly mark this answer as Accepted in case it helped or post your feedback !

Regards