According to the Microsoft Azure AI Fundamentals (AI-900) study guide and the official Microsoft Learn module "Describe features of common AI workloads", QnA Maker (now part of Azure AI Language services) allows developers to build, train, and publish a knowledge base that provides natural-language answers to user queries. A key capability of this service is active learning, which enables the knowledge base to automatically suggest improvements by analyzing user feedback and usage patterns.
Active learning is an iterative process in which the service observes real user interactions and identifies ambiguous questions or pairs of similar questions that produce uncertain or multiple answers. The system then recommends updates or refinements to the knowledge base to improve the accuracy and relevance of responses. This feedback loop helps ensure that over time, the chatbot's responses align more closely with actual user expectations and language variations.
In contrast:
* A. Key phrase extraction identifies main ideas in text and is used in content summarization, not in response optimization.
* B. Sentiment analysis detects emotional tone (positive, negative, neutral), but it doesn't refine QnA responses.
* C. Business logic defines operational rules in an application, not machine learning-driven feedback.
The AI-900 guide specifically emphasizes that QnA Maker supports active learning to improve the quality of answers based on end-user feedback, making this the verified and official Microsoft answer.
Reference (from Microsoft Learn AI-900 content):
"Active learning uses feedback from end users to automatically suggest improvements to a knowledge base, helping improve the accuracy of answers over time."