AWS AI Practitioner Question 31

Key Takeaways

AWS AI practitioner exam prep. So, question 31. A data science team is preparing training data for sentiment analysis modeling. They have 100,000 customer reviews that need to be labeled as positive, negative, or neutral. The team wants to reduce labeling costs while maintaining high accuracy. Which AWS service should the team use? So, let's identify the key hints. So, 100,000 reviews, which is a large labeling job. Labeled as positive, negative, or neutral. So, human annotation needed. Some kind of sentiment analysis, really. Reduce the labeling costs. So, they [music] want to some kind of managed workforce option. And it looks like they're going manual in this particular piece. So, option one, Amazon Comprehend custom classification with auto label. Option two, Amazon Mechanical Turk integrated directly with S3. Option three, Amazon Sagemaker Ground Truth with automated labeling workflows. Option four, Amazon Recognition custom labels for text classification. Drop your answer in the comments. So, the hint says reduced labeling costs. So, we need a service that optimizes the labeling workflow. So, option one, Comprehend custom classification requires already labeled data to train. Doesn't help you label data in the first place. Option two, Mechanical Turk is a raw crowdsourcing platform. It doesn't optimize labeling costs with automation. Option four, Recognition custom labels is for image classification, not text. It's the wrong service entirely. So, the correct answer is actually option three, Amazon Sagemaker Ground Truth. So, Ground Truth manages the full data labeling workflow. Uses active learning to auto label high confidence examples. It only sends uncertain samples to human reviewers. And it can reduce labeling costs by up to 70%. So, if you need to label training data at scale, Sagemaker's Ground Truth sub-service, it automates what it can and lets humans handle the [music] rest. Do you want to learn more about AI and AWS? Head to aws.cocould.com.