AI-102 Question 273
Select 3You are building a conversational AI solution using Azure Language Understanding (LUIS). After training your model with intents and entities, you want to evaluate its performance to ensure it meets the accuracy requirements before deployment. Which of the following actions should you take to evaluate the model's performance?
- A
Use the Test panel in the LUIS portal to test the model with sample utterances and check the prediction results.
- B
Export the model and test it using a third-party tool to measure its accuracy.
- C
Use the built-in Batch Testing feature in the LUIS portal to test the model with a dataset of labeled utterances.
- D
Deploy the model to an endpoint and capture user interactions to measure performance metrics.
- E
Analyze the Confusion Matrix and Precision/Recall scores provided by the LUIS portal after running tests.
Show answer and explanation
Correct answers: A, C, E
Explanation
Evaluating a LUIS model's performance involves using the built-in tools provided by the Azure portal, such as the Test panel for manual validation, the Batch Testing feature for systematic evaluation with labeled datasets, and analyzing performance metrics like the Confusion Matrix and Precision/Recall scores. These steps ensure that the model meets accuracy requirements before deployment. Exporting to third-party tools or deploying the model to production is not necessary for initial evaluation.
- A. Correct.
This is correct. The Test panel in the LUIS portal allows you to test the model with individual sample utterances and validate the prediction results manually.
- B. Incorrect.
This is incorrect. While exporting and using third-party tools might be an option, Azure provides built-in tools for evaluation, making this step unnecessary.
- C. Correct.
This is correct. The Batch Testing feature in LUIS allows you to upload a labeled dataset and evaluate the model's performance systematically.
- D. Incorrect.
This is incorrect. While deploying the model to capture real-world interactions is useful for monitoring in production, it is not the primary method for pre-deployment evaluation.
- E. Correct.
This is correct. The Confusion Matrix and Precision/Recall scores provided in the LUIS portal are essential tools for analyzing the model's performance during evaluation.