Hearst Newspapers: Engaging readers with cloud machine learning

For newspaper publishers, precision and speed are critical to engaging readers with the right editorial and advertising content on digital properties. At Hearst Newspapers, the newspaper division of Hearst, one of the world’s largest mass media publishers, classifying content used to be a difficult process. Sorting, labeling, and categorizing an average of 3,000 new articles every day was time consuming, and teams often had to prioritize certain content and leave other articles unclassified just to keep up.

Instead of hiring a larger team, Hearst Newspapers is solving the problem with Google Cloud AI. Using Google Cloud Natural Language API to enable content classification with powerful machine learning models in an easy-to-use REST API, Hearst Newspapers can understand what its content is about, regardless of how it is structured and presented on the company’s many websites. Although Hearst Newspapers previously used a legacy system that attempted to automate the classification process, it was not as fast or as accurate.

“Google Cloud Natural Language API is unmatched in its accuracy for content classification,” says Naveed Ahmad, Senior Director of Data at Hearst Newspapers, who is responsible for data centralization and business intelligence using Google Cloud Platform. At Hearst Newspapers, we publish several thousand articles a day across more than 30 properties. With natural language processing, we can quickly gain insight into what content is being published and how it resonates with our audiences.”

Revolutionizing personalization

With Google Cloud Platform, Hearst Newspapers is finding new ways to use machine learning to personalize content and engage readers. Google Cloud Natural Language API integrates with DoubleClick for Publishers by Google, helping the Hearst Newspaper editorial team save time previously spent manually tagging content. It also improves ad targeting to increase revenue. This allows Hearst to analyze its past revenue performance in the context of content and predict future ad performance by content categories.

With the ability to classify documents in more than 700 predefined categories—such as news, technology, health, and entertainment—Hearst Newspapers can automatically parse the meaning of articles in real time to organize and present them more efficiently. The company also uses entity recognition in Google Cloud Natural Language API to quickly identify people, places, and events featured in content.

“With the granular data we get from Google Cloud Natural Language API, we can target more specific content segments and associate ads with narrower reader profiles,” says Naveed. “We save a lot of time because we don’t have to manually categorize 3,000 articles daily.”

This capability will not only save the ad operations team several hours a day, it will also allow for a much larger content inventory for matching ads. Previously, the team manually classified only a portion of the content for better ad targeting. Now, all new content is classified automatically.

The Google Cloud Natural Language API category and entity data is also pushed to the Hearst Newspapers Customer Data Platform, where it is used to create user profiles and create segments based on reading habits. This allows for targeted messaging and offers to these segments.

Hearst Newspapers also pulls all the CMS content, along with the Google Cloud Natural Language API tags, into a Google BigQuery data warehouse. By marrying this data to Google Analytics, the company can do sophisticated analysis on the nature of the content being published and how it is consumed over time. This analysis will help drive the future strategy of how the company can improve and invest into curation of content, and better understand how readers engage.

A new world for publishers

Hearst Newspapers is exploring a broad spectrum of Google Cloud AI machine learning services, from out-of-the-box APIs to a content recommendation engine based on TensorFlow, an open-source library for machine intelligence developed by Google.

Hearst Newspapers also uses Google BigQuery to analyze content performance and calculate the value of ad slots. Google BigQuery Data Transfer Service allows the company to centralize data from Google Analytics and internal data sources, helping the company analyze its advertising campaigns against web traffic trends.

“Google Cloud Platform and cloud machine learning services open up a new world for Hearst, and for all publishers to get more value from data,” says Naveed. “We look forward to the possibilities.”