From the course: Developing Modern Applications with AWS AI and Generative AI Services

Amazon Comprehend - Amazon Web Services (AWS) Tutorial

From the course: Developing Modern Applications with AWS AI and Generative AI Services

Start my 1-month free trial Buy for my team

Amazon Comprehend

“

- [Narrator] Let's explore how to process documents and text using AWS AI services, beginning with an in-depth exploration of Amazon Comprehend. Amazon Comprehend is a sophisticated, fully managed, natural language processing service that leverages advanced machine learning algorithms to analyze, text, and extract valuable insights. This service operates seamlessly, freeing users from the burden of building or training their own models. Let's explore what Amazon Comprehend can accomplish. We will focus on three key functionalities. First, it excels at analyzing documents to extract insights such as entities, key phrases, and personally identifiable information, and determining the overall sentiment of the text. For instance, consider the input text, "Bob has been living at 123 Express Lane "since early 2000." Amazon Comprehend will identify Bob as a person, 123 Express Lane as an address, and the number 2000 as a date. Each entity is assigned a score indicating the level of confidence. Next is a custom classification feature, which empowers users to categorize their documents into predefined groups without needing a background in machine learning. This is particularly useful for organizations inundated with daily customer support emails. For example, if your business receives thousands of these emails, you can set up Amazon Comprehend to automatically categorize them into specific buckets, such as billing issues, technical support, account management, and general inquiries, facilitating efficient routing to the appropriate departments. Lastly, we have custom entity recognition, which enhances Amazon Comprehend's capabilities by allowing users to tag domain-specific entity types that fall outside the standard categories provided. Utilizing AutoML, Amazon Comprehend learns from a small set of designated examples, subsequently training a private custom model capable of recognizing these specialized terms across various types of text, including Word documents, PDFs, and plain text. For instance, if your company specializes in electronics such as smartphones, laptops, and smartwatches, you might receive plenty of customer reviews. Amazon Comprehend can be configured to automatically extract specific defect related entities. Example, hardware issues, software issues, screen damage, overheating from these reviews, enabling you to analyze prevalent compliance, and enhance product quality based on customer feedback. Having established what Amazon Comprehend can do, let's explore the operational mechanics. The initial step involves inputting your text data, which can consist of various sources, such as social media comments, emails, customer feedback, or legal documents. This data can be efficiently uploaded from a file, or streamed directly from an application into an Amazon S3 bucket. Once your text is ingested, Amazon Comprehend processes it to perform tasks such as language reduction, sentiment analysis, or document classification. The insights generated are then returned in structured formats like JSON, or CSV, which can be easily integrated with other services such as Amazon Redshift for in-depth analytics. Some common use cases for Amazon Comprehend include analyzing customer sentiment, extracting insights from surveys, safeguarding sensitive legal documents through PII protection, and facilitating fraud detection efforts. Each of these applications demonstrates the power of Amazon Comprehend in processing and extracting meaningful insights from textual data.

Contents