AI-102 Question 312
Select 3You are building a QnA Maker knowledge base for a customer support chatbot using Azure AI services. After importing a FAQ document into the knowledge base, you notice that the responses to user queries are not accurate during testing. What steps should you take to improve the accuracy of the knowledge base?
- A
Re-train the knowledge base after manually editing question-response pairs.
- B
Add alternative user questions (utterances) for existing question-response pairs.
- C
Increase the pricing tier of the QnA Maker service to improve AI accuracy.
- D
Enable active learning and review user feedback to refine the knowledge base.
- E
Delete the knowledge base and start over with a new FAQ document.
Show answer and explanation
Correct answers: A, B, D
Explanation
To improve the accuracy of a QnA Maker knowledge base, you should focus on refining the question-response pairs, adding alternative user utterances, and leveraging active learning to incorporate user feedback. These steps help the model better understand user queries and provide accurate responses. Upgrading the service tier or deleting the knowledge base are not appropriate actions for this scenario.
- A. Correct.
Manually editing question-response pairs and re-training helps improve response accuracy by ensuring the knowledge base understands the intent behind user inputs.
- B. Correct.
Adding alternative user questions (utterances) improves the model's ability to match variations of user queries to the correct response.
- C. Incorrect.
Increasing the pricing tier does not directly impact the accuracy of the knowledge base; it only changes performance characteristics like scaling and latency.
- D. Correct.
Enabling active learning allows the system to suggest potential improvements based on user interactions, which can refine the knowledge base.
- E. Incorrect.
Deleting the knowledge base is unnecessary and inefficient; improvements can be made incrementally without starting over.