Custom Question Answering – Fundamentals of Natural Language Processing

Custom Question Answering

QnA Maker is now available in Azure Portal as a custom question-answering service under the Language Service category. You can create a knowledge base out of question-and-answer pairs by using the Language Studio’s custom question-answering feature. You can get these questions and answers from an existing FAQ document or website, or you can type them in and edit them by hand. Most of the time, all of these methods are used together to make a knowledge base, starting with a source dataset of questions and answers from an existing FAQ document and adding more manual entries to the knowledge base. You can also assign an alternate phrase to the questions in the knowledge base. For example, “What is your school name?” may have the alternate phrase “In which school do you study?”

Test the Knowledge Base

You must save a set of question-and-answer pairs after you have created it. This process looks at your literal questions and answers and uses a built-in NLP model to match the right answers to your questions, even if the answers aren’t exactly worded the way you said they should be in your question definitions. Then, you can use Language Studio’s built-in test interface to see how much you know by asking questions and reading the answers.

Extend and Customize the Bot

• After you’ve set up your bot, you can change it by going to the Azure portal, where you can: Use an appropriate testing interface to test your bot.
• Logging, analytics, and interaction with other services should be configured.
• To make it more advanced, rewrite the code and release the updated version of the bot.

You can edit both codes at once within the Azure portal for brief modifications. But in case you want to make more modifications, you may download the source code and manipulate it on your own computer before republishing the bot without delay to Azure.