Applications of Foundation Models
AWS AI Services
AWS AI services are managed, task-specific services that often avoid custom model training. The exam commonly asks you to match a use case to Comprehend, Rekognition, Textract, Polly, Transcribe, Translate, Kendra, Personalize, Lex, Forecast, Fraud Detector, Lookout services, or A2I.
Concepts
- Amazon Comprehend extracts insights from text such as sentiment, entities, key phrases, language, and topics.
- Amazon Rekognition analyzes images and videos for labels, moderation, faces, text, and custom labels.
- Amazon Textract extracts text, handwriting, forms, and tables from documents.
- Amazon Polly turns text into lifelike speech; Amazon Transcribe turns speech into text; Amazon Translate translates text.
- Amazon Kendra provides enterprise semantic search; Amazon Personalize builds recommendations; Amazon Lex builds conversational bots.
- Lookout services address anomaly detection for equipment, metrics, and vision; Amazon Fraud Detector targets online fraud; Amazon Forecast provides time-series forecasting; Amazon A2I adds human review.
Exam tips
- Use Comprehend for NLP, Rekognition for images/video, and Textract for document extraction.
- Use Polly for text-to-speech, Transcribe for speech-to-text, and Translate for translation.
- Use Personalize for recommendations, Kendra for enterprise search, and Lex for conversational bots.