⬡ Hub
Skip to content

AWS AI Services: Specialized Managed Services

Beyond SageMaker and Bedrock, AWS offers a suite of "AI Services" which are pre-trained models accessible via simple API calls. No machine learning expertise is required.

1. Natural Language & Text

📝 Amazon Comprehend

  • Purpose: NLP to extract insights and relationships in text.
  • Features: Sentiment Analysis, Entity Recognition, Keyphrase Extraction, Medical (Specialized).
  • Example: comprehend.detect_sentiment(Text="I love AWS!", LanguageCode="en")

📄 Amazon Textract

  • Purpose: Automatically extracts text, handwriting, and data from scanned documents.
  • Features: Form recognition (key-value pairs), Table extraction, Query-based extraction.

2. Speech & Language Translation

🗣️ Amazon Polly

  • Purpose: Text-to-Speech (TTS). Converts text into life-like speech.
  • Features: Neural TTS (Highly natural), SSML support (control prosody), Lexicons.
  • Example: polly.synthesize_speech(Text="Hello!", OutputFormat="mp3", VoiceId="Joanna")

🎤 Amazon Transcribe

  • Purpose: Automatic Speech Recognition (ASR). Speech-to-Text.
  • Features: Real-time streaming, Speaker identification (diarization), PII redaction (Call centers).

🌐 Amazon Translate

  • Purpose: Neural machine translation service.
  • Features: Real-time translation, Batch translation, Custom terminology.

3. Specialized Business Services

Service Category Use Case
Personalize Recommendation Build real-time recommendations (like Amazon.com).
Lex Chatbots Build conversational interfaces (The technology behind Alexa).
Kendra Search Enterprise search service powered by ML (finds answers in docs).
Forecast Time Series Predict future business metrics (inventory, revenue).
Fraud Detector Security Identify potentially fraudulent online activities.

4. Technical Strategy: Which Service When?

  1. Need to recognize a person? -> Rekognition (Collections).
  2. Need to extract data from an Invoice? -> Textract.
  3. Need to search through a massive PDF repository? -> Kendra.
  4. Need to analyze customer sentiment on social media? -> Comprehend.
  5. Need to build your own custom LLM or use GPT-like models? -> Bedrock / SageMaker.

📝 Implementation Snippet: Sentiment Analysis (Boto3)

import boto3

comprehend = boto3.client('comprehend')
text = "The product arrived early and works perfectly!"

response = comprehend.detect_sentiment(Text=text, LanguageCode='en')
print(f"Sentiment: {response['Sentiment']}") # Output: POSITIVE