AI-102 Question 269
Select 3You are designing a language understanding (LUIS) model for an e-commerce application that allows customers to search for products and check order statuses using natural language. After training the model with appropriate intents and entities, you need to evaluate its performance. Which of the following actions should you take during the evaluation phase to ensure the model performs accurately?
- A
Use real-world user utterances to test the model.
- B
Manually inspect and correct misclassified intents in the test data.
- C
Rely exclusively on the training dataset for evaluation.
- D
Analyze the model's precision, recall, and F1 score.
- E
Deploy the model before evaluating its performance.
Show answer and explanation
Correct answers: A, B, D
Explanation
The evaluation phase is crucial to ensure the LUIS model performs accurately before deployment. It involves testing with real-world data, analyzing performance metrics, and addressing misclassifications. These steps help refine the model and ensure it generalizes well to unseen user inputs.
- A. Correct.
Using real-world user utterances ensures the model is tested under realistic conditions and helps identify issues that may not be apparent from training data alone.
- B. Correct.
Manually inspecting and correcting misclassified intents allows you to refine the model and improve its accuracy during evaluation.
- C. Incorrect.
Relying exclusively on the training dataset can lead to overfitting and does not accurately measure the model's performance on unseen data.
- D. Correct.
Analyzing precision, recall, and F1 score provides key performance metrics to understand how well the model is performing in classifying intents and extracting entities.
- E. Incorrect.
Deploying the model before evaluating its performance is not a best practice. Evaluation should occur before deployment to ensure the model is ready for production.