How Publishers Are Using Machine Learning to Improve Content Creation & Advertising
Machine learning, a type of artificial intelligence that allows computers to learn without being programmed, may seem like a technology pulled off the set of Star Trek, but it’s actually being used in a variety of industries. Today machine learning is used to identify and stop credit card fraud, prevent spam from entering our inbox, and improve financial investing.
Publishers too are beginning to experiment with machine learning technology. Media companies are finding applications for the technology on the editorial side, improving how reporters gather information and report the news. And some publishers, like the Associated Press, are even employing machine learning and natural language generation to write the news. There are also uses for artificial intelligence on the advertising side. Computers can learn about reader behavior, based on the first-party data publishers collect, pinpoint where they are in the buying process, and learn how to deliver the most effective messaging to them.
Key to both of these strategies is automation. Computers can ingest massive amounts of data and quickly learn what that data means and even determine the best course of action. Machine learning allows publishers to be more efficient and focus on the most important sides of their business – creating content and driving new revenue.
Following are some of the ways publishers have implemented machine learning.
Streamlining the Newsroom
In 2015 The New York Times Research and Development Lab announced an experimental project called Editor, which had the goal of simplifying the reporting process. The platform uses machine learning to identify and categorize parts of a story as it is being written. Editor identifies people, places, organizations, and events, and understands the difference between, say, George Washington the person and George Washington the university. A journalist can further tag her article to call out the headline, byline, and main point of the piece. Over time, these semantic tags will teach the computer to recognize the important parts of an article.
In an interview with Poynter, creative director of the NYT lab Alexis Lloyd explained the goal of Editor: “I think there could be all sorts of microservices that service journalism. You could imaging microservices that could do things like try to identify quotes from people and ones that could try to find relationships between people and organizations in the text.”
Services like Editor can simplify research and fact checking by gathering all coverage of a certain person or organization in one spot. See the video below to see the Editor platform in action.
BBC News Lab launched a similar tagging technology in 2015 called “Juicer”. The difference between the two is that the Juicer platform doesn’t tag articles in real-time, and it relates tags to content outside of the BBC universe. So journalists can quickly find all of the latest stories about Donald Trump from across the web, for example.
This type of technology can also be put to use to improve the user’s experience, noted Jaqui Maher, interactive journalist at the BBC News Lab, in an interview with Poynter. BBC is already looking into ways to have related information pop up when users hover over certain keywords. The media company also wants to bring this capability to its audio and video content. Maher told Poynter that this could look like VH1’s “Pop-Up Video,” which took music videos and overlaid interesting facts on different parts of the shot. The difference being, machine learning can automate that process.
Predicting Consumer Purchases
Another way that publishers can implement machine learning is on the sales side of the business. Recently, Condé Nast announced a major push into the machine learning space with the launch of “Condé Nast Spire,” a data offering that identifies what consumers are interested in and where they are in the buying cycle. Condé Nast’s first-party behavioral data combines with consumer purchase data from data solutions company 1010data (owned by Condé’s parent company Advance Publications) to identify specific segments of the audience. Then the platform uses machine-learning techniques to optimize advertising messages to these segments and do more of what works.
“We are again moving the industry forward by giving our advertising partners the ability to optimize campaigns in real time through the strategic use of our extensive data capabilities,” said Edward Menicheschi, chief marketing officer at Condé Nast in a press release.
While still in beta, Condé Nast Spire identified that a segment of users that consumed more humor, design, and politics content in video format where more likely to purchase computers. When Spire served computer advertisements and reviews to that group, computer purchases were made 25% sooner.
A Gamechanger for Content & Sales
Machine learning is creating new ways for publishers to understand the data that they collect and streamline manual processes into automated computer operations. That is opening up new opportunities in content creation and monetization. The above stories are just a few examples, but there are and will be many more applications for this advanced technology. Machine learning is a space publishers must watch as it will impact all parts of the business.