QnA Maker is a cloud-based Natural Language Processing (NLP) service that easily creates a natural conversational layer over your data. It can be used to find the most appropriate answer for any given natural language input, from your custom knowledge base (KB) of information.
At its initial launch the capabilities of QnA Maker were limited to single turn questions in which one question relates to one answer. But some questions cannot be answered in a single turn. At Build, Microsoft announced a set of new features including multi-turn support with dialogs and follow-up questions to refine the search for a final answer.
In this session, we will cover the lessons learned of building an enterprise chatbots with multiple knowledge bases consisting of multi-turn questions. We will start with the conversation design, but quickly move to a multi-turn chatbot and explain how a dialog can be created by using the multi-turn functionality. We then discuss how to use meta data, labelling the questions and answers to filter and refine the search in certain topics. Once we have trained, tested and published the knowledge base- we will wire it up by using the Microsoft Bot Framework to include:
- Logic required for supporting multi-turn
- Logic for dispatch to support multiple knowledge bases by using LUIS
- Logic to cancel the conversation at any moment
- Logic to add meta data for testing purpose
- Logic to implement the feedback loop by using Application Insights
- Logic to dump the full conversation for later analyses
Benefits of Attending one of these sessions
- Complete overview of Azure Cognitive Services
- Better understand the use cases by the packed demos
- Understand when to use the strengths and weak points
- Learn how to build multi-turn chatbots with QnA Maker
- How to implement the feedback loop for conversational interfaces