SAA-C03 ยท Durability, automation and observability
Purpose-built managed services
Raising reliability without code changes by using managed services (RDS, EFS, ELB, Amazon MQ) and managed AI services such as Comprehend, Polly, Rekognition and Transcribe.
Key points
- Swap self-managed parts for managed ones, often with no code changes.
- DB on one EC2 instance -> RDS Multi-AZ (automatic failover, same endpoint); NFS share -> Amazon EFS (across AZs).
- Single web server -> Elastic Load Balancing with health checks; ActiveMQ or RabbitMQ -> Amazon MQ (same protocols).
- AI: Comprehend (NLP: sentiment, entities, key phrases, PII), Polly (text to speech), Transcribe (speech to text), Rekognition (image, video).
Source: https://docs.aws.amazon.com/comprehend/latest/dg/what-is.html
Practice questions
Which AWS service detects the sentiment, entities and key phrases in customer reviews using pre-trained NLP models?
- Amazon Polly
- Amazon Comprehend
- Amazon Transcribe
- Amazon Rekognition
Show answer
Answer: B. Amazon Comprehend
Amazon Comprehend is the managed natural language processing service for sentiment, entities, key phrases, language and PII. Polly turns text into speech, Transcribe turns speech into text, and Rekognition analyzes images and video.
Which pairings of task and managed AI service are correct?
- Turn recorded support calls into text: Amazon Transcribe
- Read news articles aloud in an app: Amazon Polly
- Detect objects in uploaded photos: Amazon Rekognition
- Convert audio files directly into text: Amazon Comprehend
- Translate a PDF contract into German: Amazon Polly
Show answer
Answer: A. Turn recorded support calls into text: Amazon Transcribe, B. Read news articles aloud in an app: Amazon Polly, C. Detect objects in uploaded photos: Amazon Rekognition
Transcribe converts speech to text, Polly produces lifelike speech from text, and Rekognition analyzes images and video, for example to find objects or faces. Comprehend works on text, not audio, and Polly does not translate documents.
To use Amazon Comprehend for sentiment analysis, you must first collect training data and train your own machine learning model.
Show answer
Answer: False
False. Comprehend's insights such as sentiment, entities and key phrases use pre-trained models, so no training data is needed. Custom training is optional and only for your own classifiers or entity types.
Practice 8 questions on this topic in Bitzu
Short lessons, instant explanations and spaced review of what you get wrong. Free on iOS and Android.
