Amazon Comprehend

Amazon Comprehend

Amazon Comprehend is a natural-language processing (NLP) service that uses machine learning to uncover valuable insights and connections in text. Comprehend allows you to harness the power of machine learning:

  1. differentiate your business with the automation of classifying documents and identifying key terms
  2. simplify workflows and document processing by extracting key phrases, customer sentiment and topics
  3. protect and secure sensitive information by identifying and redacting Personal Identifying Information (PII) from documents

Top 5 things to know about Amazon Comprehend

  1. It's an API based system
  2. It works with written content
  3. It's designed to operate on a massive scale
  4. It's configured and managed through the AWS console
  5. It can be customized

1. Amazon Comprehend is an API based system

There is no graphical user interface (GUI) for Comprehend. Data is uploaded via an API service and responses from Comprehend comes in the form of a JSON array response. If this terminology doesn't make sense to you, you can still benefit from Comprehend, but you'll need technical help. Contact us today for a free estimate.

2. Amazon Comprehend works with written content

If you want to use some of the features of Amazon Comprehend with content like audio or video, you'll need to convert that to written text first. Amazon Transcribe is a service that does exactly that.

3. Amazon Comprehend is scalable

Amazon Comprehend is designed to work with millions of documents, though it can also be used on a much smaller scale.

4. Amazon Comprehend configured and managed through the AWS console

Like all AWS products and services, the first step is getting setup with an AWS account and logging into the AWS Console. If you are feeling overwhelmed once you've logged in, please feel free to reach out and we'd be happy to help.

5. Amazon Comprehend can be customized

Out of the box Amazon Comprehend can parse and create rich metadata for many types of content. However if you have specialized data, the default deep learning algorithm can be customized to fit. There's even an Amazon Comprehend Medical, and there may be more specialized algorithms to come.

Amazon Comprehend Features

Specific Amazon Comprehend features include the following:

  • Key phrase extraction - returns the key phrases or talking points and a confidence score.
  • Sentiment analysis - returns the overall sentiment of a text (Positive, Negative, Neutral, or Mixed).
  • Targeted sentiment - more granular sentiment insights by identifying the sentiment (positive, negative, neutral, or mixed) towards entities within text.
  • Entity Recognition - returns the named entities —people, places, things— which are automatically categorized.
  • Language Detection - identifies text written in over 100 languages and returns the dominant language with a confidence score.
  • Event detection - extracts the event structure from a document, distilling pages of text down to easily processed data for consumption by AI applications or graph visualization tools.
  • Syntax Detection - analyzes text using tokenization and Parts of Speech (PoS), and identify word boundaries and labels like nouns and adjectives within the text.
  • Detect PII - detect and redact personally identifiable information (name, social security numbers, emails, etc)

Amazon Comprehend use cases

  1. Call Center Analytics & Data Mining - Analyze customer interactions, understand customer sentiment and automatically categorize inbound calls. Gain insight from customer surveys to improve your business.
  2. Product Review Indexing - Focus on context by equipping your search engine to index key phrases, entities, and sentiment, not just keywords.
  3. Legal Briefs - Extract insights automatically from packets of legal briefs such as contracts and court documents. Enhance security by identifying and redacting Personally Identifiable Information (PII).
  4. Financial Documents - Classify and documents such as insurance claims or mortgage packages or find relationships between financial events in a financial article.

How much does Amazon Comprehend Cost?

Pricing for Amazon Comprehend is done on a per unit basis, meaning you only pay for the times you utilize the API.  Units are 100 characters with a 3 unit (300 character) minimum charge per request. Individual features are charged on a per unit basis, with most features costing 1/100th of a penny per unit.  Event detection cost more; syntax and PII detection costs less.

Bulk Discounts

Processing more than 10 million units ($100/feature) means that prices go even lower on a per unit basis, essentially being cut in half.

Free Tier

If you don't need to process a ton of information, you might be able to use the free tier, which is anything under than 50,000 units (see above) or 5 million characters per month. Eligible features include:

  • Key Phrase Extraction
  • Sentiment
  • Targeted Sentiment
  • Entity Recognition
  • Language Detection
  • Event Detection
  • Syntax Analysis
  • Detect/Contains PII

Complete detailed pricing can be found here.

Customizing Amazon Comprehend

You can also use Comprehend to train a custom NLP model to categorize text and extract custom entities. Asynchronous inference requests are measured in units of 100 characters, with a 3 unit (300 character) minimum charge per request. You are charged $3 per hour for model training (billed by the second) and $0.50 per month for custom model management. For synchronous Custom Classification and Entities inference requests, you provision an endpoint with the appropriate throughput. You are charged from the time that you start your endpoint until it is deleted.

Complete documentation on custom entity recognition, including documentation on preparing the training data, training recognizer models, and running custom recognizers can be found here.

Need Help?

Need help with an Amazon Comprehend project? Please don't hesitate to reach out for a free consultation.

Learn more about how we've set our clients up for success by reading our case studies

Have questions?