MLA-C01 Practice Question 94: Comprehend Custom Classification
AWS Certified Machine Learning Engineer – Associate (MLA-C01) · Test 10 · question 4 of 10
An insurance company receives thousands of customer emails a day and wants to route each one to one of eight internal queues, such as claims, billing, and policy changes. It has 25,000 past emails already labeled with the correct queue. The company has no data scientists and wants a managed AWS service that learns its categories from the labeled emails and classifies new emails through an API.
Which approach should the company use?
Show answer and explanation
Answer:
Amazon Comprehend custom classification sorts documents into categories you define. You first train a custom classifier on labeled examples, then classify documents one at a time in real time or in batches with an asynchronous job. The company's 25,000 labeled emails are the training data, the eight queues are the classes, and the service handles training and hosting without data scientists (option A).
- B is wrong because hand-written rules over key phrases are brittle and ignore the labeled examples. Keeping rules accurate for eight categories of free-form email becomes its own ongoing project. - C is wrong because topic modeling is unsupervised. It finds its own themes, which will not line up neatly with the company's eight queues, and it ignores the labels the company already has. - D is wrong because Amazon Lex builds conversational bots for live exchanges with users. It is not a document classification service for routing long emails in bulk.
Reference: Custom classification - Amazon Comprehend