Hands-On Lab: Exploring Specialized AWS AI Services

Welcome to this 30-minute guided lab! In this session, you will explore AWS's specialized AI services, specifically focusing on Natural Language Processing (NLP) with Amazon Comprehend and Text-to-Speech (TTS) with Amazon Polly. These pre-trained services allow you to add advanced AI capabilities to your applications without needing deep machine learning expertise.

Prerequisites

Before starting this lab, ensure you have the following:

• AWS Account: An active AWS account with permissions to access Amazon Comprehend and Amazon Polly.
• CLI Tools: The AWS CLI (aws) installed and configured on your local machine.
• IAM Permissions: The configured AWS profile must have ComprehendFullAccess and AmazonPollyFullAccess (or equivalent administrator rights).
• Prior Knowledge: Basic familiarity with navigating the terminal and understanding JSON structures.

Learning Objectives

By completing this lab, you will be able to:

1. Analyze unstructured text to detect sentiment using Amazon Comprehend.
2. Extract key phrases and entities from text data.
3. Convert textual data into lifelike speech using Amazon Polly.
4. Seamlessly switch between the AWS CLI and the AWS Management Console to interact with specialized AI APIs.

Architecture Overview

AWS offers a broad taxonomy of AI services. This lab focuses on the Language and Speech domain.

In this specific lab scenario, you will pass a single string of text into both services to observe their different capabilities:

Step-by-Step Instructions

Step 1: Detect Sentiment with Amazon Comprehend

Amazon Comprehend uses NLP to extract insights from text. It returns a dominant sentiment along with confidence probability scores $P(\text{Sentiment} | \text{Text}) \in [0, 1]$ for Positive, Negative, Neutral, and Mixed.

Run the following command in your terminal to analyze a sample review:

[!TIP] Notice the output JSON. Look for the Sentiment key and the SentimentScore object to see how confident the AI is in its assessment.

▶💻 Console Alternative

1. Log into the AWS Management Console.
2. Search for Amazon Comprehend and select it.
3. On the left sidebar, click Real-time analysis.
4. Scroll down to the Input text box and paste: "I am absolutely thrilled with the new machine learning features in AWS. They are incredibly easy to use!"
5. Click Analyze.
6. View the results under the Insights -> Sentiment tab.

📸 Screenshot: Real-time analysis sentiment results showing "Positive".

Step 2: Extract Key Entities

Comprehend can also pull out named entities (like Organizations, Locations, and People).

▶💻 Console Alternative

1. In the same Real-time analysis page from Step 1, paste the text: "Amazon Web Services is located in Seattle, Washington."
2. Click Analyze.
3. Click the Entities tab below to see "Amazon Web Services" classified as an ORGANIZATION and "Seattle, Washington" as a LOCATION.

Step 3: Synthesize Speech with Amazon Polly

Now, let's take text and convert it to lifelike speech using Amazon Polly.

[!TIP] You can change the --voice-id to Matthew or Salli for different voices.

▶💻 Console Alternative

1. In the AWS Management Console, search for Amazon Polly.
2. In the left navigation pane, choose Text-to-Speech.
3. Under the Text section, paste the same sentence.
4. Select a voice (e.g., Joanna, Female).
5. Click Listen to speech to play it directly in your browser, or click Download to save the MP3.

📸 Screenshot: Amazon Polly Text-to-Speech interface.

Checkpoints

Verify that your lab steps were successful by performing these checks:

1. Verify Comprehend Output: Run aws comprehend detect-sentiment from Step 1 again. Does the output show "Sentiment": "POSITIVE" with a confidence score greater than 0.95?
2. Verify Polly Output: In your terminal, verify that the MP3 file was created:

(Expected outcome: You should see output-speech.mp3 listed with a file size greater than 0 bytes. Try playing it using your computer's media player!)

Clean-Up / Teardown

[!WARNING] Remember to run the teardown commands to avoid ongoing clutter, though note that Amazon Comprehend and Amazon Polly real-time API calls do not provision ongoing infrastructure. You are only billed per request (character count).

Since this lab only utilized stateless API calls, there are no EC2 instances or SageMaker endpoints to delete. However, you should clean up your local workspace:

Troubleshooting

If you run into issues during the lab, reference this troubleshooting table: