Amazon Comprehend

Amazon Comprehend

A Natural Language Processing (NLP) service that uses ML to find insights and relationships in text.

Key Capabilities

• Sentiment Analysis: Determine if text is positive, negative, neutral, or mixed
• Entity Detection: Identify entities (people, places, organizations, dates, quantities)
• Key Phrase Extraction: Extract important phrases from text
• Language Detection: Identify the dominant language of text (supports 100+ languages)
• Topic Modeling: Discover topics across a collection of documents
• PII Detection: Find and redact personally identifiable information in text
• Syntax Analysis: Parse sentence structure (parts of speech, tokens)
• Targeted Sentiment: Sentiment analysis for specific entities in text

Custom Classification

• Train a custom text classifier using your own labeled data
• No ML expertise required — provide labeled examples, Comprehend trains the model
• Use Cases: Custom document classification, ticket routing, content categorization

Named Entity Recognition (NER)

• Built-in recognition of standard entity types (person, location, organization, date, etc.)
• Custom Entity Recognition: Train custom NER models to detect domain-specific entities (e.g., product codes, medical terms)

Comprehend Medical

A specialized version of Comprehend designed for medical and healthcare text.

Key Capabilities

• Extract medical entities (conditions, medications, dosages, procedures, anatomy)
• Identify relationships between entities (medication-dosage, condition-treatment)
• Detect clinical trial eligibility criteria
• Process unstructured clinical notes, discharge summaries, medical records

Protected Health Information (PHI) Detection

• Automatically detect and identify Protected Health Information (PHI) in medical text
• PHI includes: Patient names, addresses, dates, medical record numbers, Social Security numbers
• Enables HIPAA compliance by identifying PHI for redaction or de-identification