Data Science Investment Counter —Funding raised by UK data science companies in 2018.
£ 5.640 Million

Amazon Comprehend adds new features to machine learning tool

Amazon announced yesterday an enhancement to its Comprehend platform that allows developers to build lists of specialised words and phrases without machine learning domain knowledge.

Comprehend is a natural language processing tool to help companies extract specific words and phrases from documents or websites.

Commenting on the upgrade, Matt Wood, GM for deep learning and AI wrote in a blog post: “Today we are excited to bring new customisation features to Comprehend, which allow developers to extend Comprehend to identify natural language terms and classify text which is specialised to their team, business or industry.”

The aim of this update, said Wood, is to let the algorithm process all the complex tasks in the background, without developers having to know the deep machine learning or natural language processing operating in the background.

“Under the hood, Comprehend will do the heavy lifting to build, train, and host the customised machine learning models, and make those models available through a private API,” Wood wrote.

The program will operate in two phases. The first one will see developers defining a list of custom entities, such as a specific vocabulary or parts of a given vehicle. Amazon will learn to identify the customised language and build a tailored model based on the list.

The second phase involves customised classifying. Once a particular list has been built, the tool will allow you to structure logical lists where the terms appear.

“Through as few as 50 examples,” Wood wrote, “Comprehend will automatically train a custom classification model that can be used to categorise all your documents. You could group support emails by department, social media posts by product, or analyst reports by business unit”.

The new Comprehend features are available from today, and you can have a look at them here.


Co-working space and blog dedicated to all things data science.

Subscribe to our newsletter