Table of Contents

Preface

Section 1: Introduction to AWS AI NLP Services

Chapter 1: NLP in the Business Context and Introduction to AWS AI Services

Introducing NLP

Overcoming the challenges in building NLP solutions

Understanding why NLP is becoming mainstream

Introducing the AWS ML stack

Summary

Further reading

Chapter 2: Introducing Amazon Textract

Technical requirements

Setting up your AWS environment

Signing up for an AWS account

Creating an Amazon S3 bucket and a folder and uploading objects

Creating an Amazon SageMaker Jupyter notebook instance

Changing IAM permissions and trust relationships for the Amazon SageMaker notebook execution role

Overcoming challenges with document processing

Understanding how Amazon Textract can help

Presenting Amazon Textract's product features

Uploading sample document(s)

Raw text or text extraction

Form data and key/value pairs

Table extraction

Multiple language support

Handwriting detection

Human in the loop

Using Amazon Textract with your applications

Textract APIs

Textract API demo with a Jupyter notebook

Building applications using Amazon Textract APIs

Summary

Chapter 3: Introducing Amazon Comprehend

Technical requirements

Understanding Amazon Comprehend and Amazon Comprehend Medical

Challenges associated with setting up ML preprocessing for NLP

Exploring the benefits of Amazon Comprehend and Comprehend Medical

Detecting insights in text using Comprehend and Comprehend Medical without preprocessing

Using these services to gain insights from OCR documents from Amazon Textract

Exploring Amazon Comprehend and Amazon Comprehend Medical product features

Discovering Amazon Comprehend

Deriving diagnoses from a doctor-patient transcript with Comprehend Medical

Using Amazon Comprehend with your applications

Architecting applications with Amazon API Gateway, AWS Lambda, and Comprehend

Summary

Section 2: Using NLP to Accelerate Business Outcomes

Chapter 4: Automating Document Processing Workflows

Technical requirements

Automating document processing workflows

Setting up compliance and control

Setting up to solve the use case

Additional IAM prerequisites

Automating documents for control and compliance

Processing real-time document workflows versus batch document workflows

Summary

Further reading

Chapter 5: Creating NLP Search

Technical requirements

Creating NLP-powered smart search indexes

Building a search solution for scanned images using Amazon Elasticsearch

Prerequisites

Uploading documents to Amazon S3

Inspecting the AWS Lambda function

Searching for and discovering data in the Kibana console

Setting up an enterprise search solution using Amazon Kendra

In this section, we will cover the steps to get started.

Walking through the solution

Searching in Amazon Kendra with enriched filters from Comprehend

Summary

Further reading

Chapter 6: Using NLP to Improve Customer Service Efficiency

Technical requirements

Introducing the customer service use case

Building an NLP solution to improve customer service

Setting up to solve the use case

Additional IAM prerequisites

Preprocessing the customer service history data

Summary

Further reading

Chapter 7: Understanding the Voice of Your Customer Analytics

Technical requirements

Challenges of setting up a text analytics solution

Setting up a Yelp review text analytics workflow

Setting up to solve the use case

Walking through the solution using Jupyter Notebook

Summary

Further reading

Chapter 8: Leveraging NLP to Monetize Your Media Content

Technical requirements

Introducing the content monetization use case

Building the NLP solution for content monetization

Setting up to solve the use case

Additional IAM prerequisites

Uploading the sample video and converting it for broadcast

Running transcription, finding topics, and creating a VAST ad tag URL

Inserting ads and testing our video

Summary

Further reading

Chapter 9: Extracting Metadata from Financial Documents

Technical requirements

Extracting metadata from financial documents

Setting up the use case

Setting up the notebook code and S3 Bucket creation

Analyzing the output of Comprehend Events

Summary

Further reading

Chapter 10: Reducing Localization Costs with Machine Translation

Technical requirements

Introducing the localization use case

Building a multi-language web page using machine translation

Setting up to solve the use case

Running the notebook

Summary

Further reading

Chapter 11: Using Chatbots for Querying Documents

Technical requirements

Introducing the chatbot use case

Creating an Amazon Kendra index with Amazon S3 as a data source

Building an Amazon Lex chatbot

Deploying the solution with AWS CloudFormation

Summary

Further reading

Chapter 12: AI and NLP in Healthcare

Technical requirements

Introducing the automated claims processing use case

Understanding how to extract and validate data from medical intake forms

Understanding clinical data with Amazon Comprehend Medical

Understanding invalid medical form processing with notifications

Understanding how to create a serverless pipeline for medical claims

Summary

Further reading

Section 3: Improving NLP Models in Production

Chapter 13: Improving the Accuracy of Document Processing Workflows

Technical requirements

The need for setting up HITL processes with document processing

Seeing the benefits of using Amazon A2I for HITL workflows

Adding human reviews to your document processing pipelines

Creating an Amazon S3 bucket

Creating a private work team in the AWS Console

Creating a human review workflow in the AWS Console

Sending the document to Amazon Textract and Amazon A2I by calling the Amazon Textract API

Summary

Further reading

Chapter 14: Auditing Named Entity Recognition Workflows

Technical requirements

Authenticating loan applications

Building the loan authentication solution

Setting up to solve the use case

Additional IAM pre-requisites

Training an Amazon Comprehend custom entity recognizer

Creating a private team for the human loop

Extracting sample document contents using Amazon Textract

Detecting entities using the Amazon Comprehend custom entity recognizer

Setting up an Amazon A2I human workflow loop

Reviewing and modifying detected entities

Retraining Comprehend custom entity recognizer

Storing decisions for downstream processing

Summary

Further reading

Chapter 15: Classifying Documents and Setting up Human in the Loop for Active Learning

Technical requirements

Using Comprehend custom classification with human in the loop for active learning

Building the document classification workflow

Setting up to solve the use case

Creating an Amazon Comprehend classification training job

Creating Amazon Comprehend real-time endpoints and testing a sample document

Setting up active learning with a Comprehend real-time endpoint using human in the loop

Summary

Further reading

Chapter 16: Improving the Accuracy of PDF Batch Processing

Technical requirements

Introducing the PDF batch processing use case

Building the solution

Setting up for the solution build

Additional IAM prerequisites

Creating a private team for the human loop

Creating an Amazon S3 bucket

Extracting the registration document's contents using Amazon Textract

Setting up an Amazon A2I human workflow loop

Storing results for downstream processing

Summary

Further reading

Chapter 17: Visualizing Insights from Handwritten Content

Technical requirements

Extracting text from handwritten images

Creating the SageMaker Jupyter notebook

Additional IAM prerequisites

Creating an Amazon S3 bucket

Extracting text using Amazon Textract

Visualizing insights using Amazon QuickSight

Summary

Chapter 18: Building Secure, Reliable, and Efficient NLP Solutions

Technical requirements

Defining best practices for NLP solutions

Applying best practices for optimization

Using an AWS S3 data lake

Using AWS Glue for data processing and transformation tasks

Using Amazon SageMaker Ground Truth for annotations

Using Amazon Comprehend with PDF and Word formats directly

Enforcing least privilege access

Obfuscating sensitive data

Protecting data at rest and in transit

Using Amazon API Gateway for request throttling

Setting up auto scaling for Amazon Comprehend endpoints

Automating monitoring of custom training metrics

Using Amazon A2I to review predictions

Using Async APIs for loose coupling

Using Amazon Textract Response Parser

Persisting prediction results

Using AWS Step Function for orchestration

Using AWS CloudFormation templates

Summary

Further reading

Other Books You May Enjoy