Learn about QnA Maker
QnA Maker enables you to power a question and answer service from your semi-structured content.
One of the basic requirements in writing your own Bot service is to seed it with questions and answers. In many cases, the questions and answers already exist in content like FAQ URLs/documents, product manuals, etc. With QnA Maker, users can query your application in a natural, conversational manner. QnA Maker uses machine learning to extract relevant question-answer pairs from your content. It also uses powerful matching and ranking algorithms to provide the best possible match between the user query and the questions.
The easy-to-use graphical user interface enables you to create, manage, train and use your service without any developer experience.
- A complete no-code experience to create a FAQ bot.
- No more throttling. Pay for hosting the service and not for the number of transactions. See the pricing page for more details.
- Scale as per your needs. Choose the appropriate SKUs of the individual components that suit your scenario. See how to choose capacity for your QnA Maker service.
- Full data compliance. The data and runtime components are deployed in the user's Azure subscription and within their compliance boundary. Read below for more details.
Key QnA Maker processes
A QnA Maker provides two key services for your data:
- Extraction: Structured question-answer data is extracted from semi-structured data sources such as FAQs and product manuals. This extraction is done when creating the knowledge base. See here to learn how to create your knowledge base.
- Matching: Once your knowledge base has been trained and tested, you publish it. This enables an endpoint to your QnA Maker knowledge base, which you can then use in your bot or app. This endpoint accepts a user question and responds with the best question/answer match in the knowledge base, along with a confidence score for the match.
QnA Maker architecture
The QnA Maker stack consists of the following parts:
QnA Maker management services (control plane): The management experience for a QnA Maker knowledge base, which includes creation, update, training, and publishing. These activities can be done through the portal or the management APIs. The management services talk to the runtime component below.
QnA Maker runtime (data plane): The data and runtime are deployed in the user's Azure subscription in a region of their choosing. Customer question/answer content is stored in Azure Search, and the runtime is deployed as as App service. Optionally, you can also choose to deploy an Application insights resource for analytics.