Machine Learning

Question 1

Amazon Rekognition

Answer 1

• Find objects, people, text, scenes in images using ML
• Facial analysis and search (user verification, people counting)
• Database of faces
• • labeling, content moderation, text detection, face detection and analysis (gender, age, emotion) face search, celebrity recognition, pathing

Question 2

Amazon Transcribe

Answer 2

• Speech to text
• ASR (automatic speech recognition)
• Automatically remove PII using redaction
• Automatic language identification
• Use case
• • transcribe calls
• • auto captioning and subtitling
• • generate metadata for media assets for search

Question 3

Amazon Polly

Answer 3

• text to speech *
• applications that talk

Question 4

Amazon Translate

Answer 4

• Natural and accurate language translation
• localize content for international users

Question 5

Amazon Lex & Connect

Answer 5

• Lex - powers Alexa
• • ASR (speech rec) to text
• • Natural language - understands intent
• • chat or call center bots
• Amazon Connect
• • Virtual contact center to receive calls and create flows
• • Integrates with CRMs
• • no upfront payments, 80% cheaper than other solutions

Question 6

Amazon Comprehend

Answer 6

• NLP (natural language processing)
• Managed, serverless service
• ML to find insights and relationships
• • language
• • extract key phrases, places, people, brands, events
• • understands positive and negative sentiment
• • analyzed text using tokenization and parts of speech
• • Automatically organizes text files by topic
• use cases
• • analyze customer interactions for postiive or negative experiences
• • create groups articles by topics

Question 7

Amazon SageMaker

Answer 7

• Fully managed machine learning model builder
• Typically needed to orchestrate a lot of servers and stuff
• ML process: predict exam score
• • get data from all students (years of experience with IT,aws, etc.)
• • get scores from exam
• • label first, then build the model (predicting score from data)
• • train and tune over time
• then deploy the model

Question 8

Amazon Forecast

Answer 8

• fully managed service with ML
• Accurate forecasts, like future sales of an item
• product demand planning, financial planning, etc.
• requires historical time series data of product features, prices, discounts, site traffic, etc.

Question 9

Amazon Kendra

Answer 9

• document search service
• extract answers from a document
• • text, pdf, html, PP, MS Word, FAQs, etc.)
• • natural language search capabilities
• • (i.e., where is the IT support desk)
• Learns from user interactions and feedback (incremental learning)
• Ability to fine-tune search results (freshness, relevance, etc.)

Question 10

Amazon Personalize

Answer 10

• Fully managed ML service to build apps with recommendations
• Like product reocmmendations, reranking, customize direct marketing
• same tech as amazon.com
• like I bought a drill so I see drill shit
• integrates into existing websites, applications, SMS, etc.
• Implement in days, not months
• retail stores, media, entertainment

Question 11

Amazon Textract

Answer 11

• Automatically extracts text, handwriting, and data from any scanned documents using AI and ML
• like a drivers license
• Extract data from forms and tables
• Read PDFs, images, etc.