Beyond the Keywords

Amazon Comprehend is a Natural Language Processing (NLP) service that uses machine learning to find insights and relationships in a body of text. Unlike a simple "Ctrl+F" search, Comprehend understands the structure and sentiment of what is written.

For the AWS Certified AI Practitioner exam, you must understand how Comprehend differs from Generative AI (it is Analytical, not Generative) and its specialized variations.

---

1. Core Competencies: The NLP Toolkit

Sentiment Analysis

Comprehend can tell you if a review or social media post is Positive, Negative, Neutral, or Mixed.

• Use Case: A customer service team automatically flagging "Negative" support tickets for immediate manager review.

Entity Recognition

It can find people, places, dates, and brand names in unstructured text.

• Use Case: An insurance company automatically pulling out "Date of Incident" and "Location" from a 10-page handwritten claim description.

Key Phrase Extraction

Identifies the "Main points" of a document.

• Use Case: A news site automatically generating tags (e.g., #Economics, #Breaking, #London) for an article.

PII Redaction

Comprehend can identify Personally Identifiable Information (PII) like social security numbers, credit card numbers, and home addresses, and "mask" or remove them.

• Use Case: Anonymizing customer data before sharing it with an external research group to ensure compliance with privacy laws.

---

2. Specialized Case: Amazon Comprehend Medical

In the medical world, words have very specific meanings. "Aspirin" isn't just an "Entity"; it's a "Medication."

Comprehend Medical is specifically trained to understand:

• Dosages: (500mg).
• Medical Conditions: (Type 2 Diabetes).
• Procedures: (MRI).
• Protected Health Information (PHI): Ensuring HIPAA compliance by identifying sensitive patient data.

---

3. Sentiment vs. Intent

Don't confuse Comprehend with Amazon Lex.

• Comprehend: Analyzes text to tell you how the person feels (Sentiment) or what the text contains (Entities).
• Lex: A service for building Conversational Bots (Chatbots).

---

4. Visualizing the NLP Pipeline

graph LR
    subgraph Data_Source
    A[Customer Email / Tweet]
    end
    
    subgraph Comprehend_Analysis
    B[Language Detection]
    C[Sentiment Analysis]
    D[Entity Extraction: Names/Dates]
    E[PII Redaction]
    end
    
    subgraph Operations
    F[Route to Manager if Negative]
    G[Populate Database Fields]
    H[Clean Privacy-safe logs]
    end
    
    A --> B
    B --> C & D & E
    C --> F
    D --> G
    E --> H

---

5. Summary: High-Scale Text Analysis

If an exam question mentions "Finding patterns in reviews," "Identifying brand names in news," or "Redacting customer credit cards," the answer is Amazon Comprehend.

---

Exercise: Identify the NLP Task

A travel company has 1 million customer reviews. They want to know:

1. Which hotels have the most "Negative" reviews?
2. Which specific brand names (e.g., "Nike", "Starbucks") are mentioned most often in the reviews for their beach locations?

Which two Comprehend features satisfy these?

• A. Image Analysis and PII Redaction.
• B. Sentiment Analysis and Entity Recognition.
• C. Custom Labels and Translation.
• D. Text-to-Speech and Video Analysis.

The Answer is B! Sentiment for the "feeling" and Entity Recognition for the "Brand Names."

---

Knowledge Check

What's Next?

We can see and we can read. Now, let’s learn to hear and speak. In our next lesson, we break down Amazon Transcribe and Amazon Polly.