Evaluate NLP Classification Models
Company: Amazon
Role: Data Scientist
Category: Machine Learning
Difficulty: easy
Interview Round: Onsite
You are interviewing for a Data Scientist internship and discussing an NLP classification project, such as classifying customer messages, search queries, or support tickets into categories. Answer the following machine learning questions:
1. Explain a confusion matrix to a group of high school students.
2. Define precision, recall, F1 score, and AUC.
3. When would you prioritize precision over recall, and when would you prioritize recall over precision?
4. How would you evaluate a multi-class classification model?
5. What is cross-entropy loss, and why is it commonly used for classification?
6. In what situations can human evaluation be better than using an automatic objective function or metric?
7. If asked to describe an NLP project you have done, what technical details and tradeoffs should you discuss?
Quick Answer: This question evaluates a candidate's competency in NLP classification model evaluation, covering confusion matrices, precision/recall/F1/AUC, cross-entropy loss, human versus automatic evaluation, multi-class assessment, and the ability to articulate technical trade-offs in an NLP project.
