Back to Basics: What is Semantic Classification?

September 23, 2016

In our ongoing “Back to Basics” series, Lotame aims to define some of the AdTech jargon that is out there. Today we are defining “Semantic Classification” – what it is, why it is used, and how it can help publishers. 

What is Semantic Classification?

Semantic analysis and classification, as it relates to our world of advertising and marketing technologies, are tools to help identify and classify the key concepts and context of a piece of content, usually an article, using machine learning and predictive analytics.

The automatic categorization of digital content, based on keywords or phrases, is here and has been for several years. Lotame launched a semantic classification solution in late 2014 to empower our clients to quickly and easily classify their content without the need to manually categorize content based on article metadata such as URL structure, breadcrumbs, or editorial tags and keywords.

Why Do I Need Semantic Classification?

Relying on metadata is often used  for editorial purposes – for example, to better understand reader viewership patterns and content consumption across a publisher’s property or properties, or to power a publisher’s proprietary search functionality. Other article metadata, like breadcrumbs and site maps, help users to navigate to the most important or interesting stories.

Article metadata, however, is not typically thought of in terms of monetizing a publisher’s user base. You see, while editorial keywords and breadcrumbs help with content categorization and navigation on a website, they were never intended to be used as a mechanism to better understand user affinities and interests for better, more relevant advertising.

Article metadata can and often does stand in as a means to understand user interests for advertising, but since it was not intended for such purposes, there are a few things to keep in mind. For starters, a human being needs to manually tag keywords to each article, which may or may not reflect the concepts advertisers are interested in reaching. As a publisher, you likely have a team of editors tagging their articles as they see relevant for that piece of content which likely introduces consistency issues. Of course, any time a human is involved in creating keywords (or using those keywords for categorization in the DMP) there is a risk of error, resulting in degraded ad performance or mismatched user interests.

Request A Demo

How Does Semantic Classification Help Me?

Instead of relying on human-created keywords and humans classifying those keywords into IAB-standard hierarchies, semantic classification reviews the actual content of an article – not the metadata – and identifies the important words and phrases taken directly from the article itself.

Importantly, relationships between words and phrases are created to other words and phrases from the same article to determine the proper context of each word or phrase. Remember, words can have many meanings. Take the word “orange” for example. Orange could refer to a fruit, a color, a bank, a band, a film, a war, a piece of software, dozens of cities and municipalities around the world, and a lot more. How do you know which orange “orange” really means? Semantic Classification tools review all content in an article to create relationships, so other words will offer clues as to which “orange” the article is discussing. Does the article mention other fruits? Other bands or films? It’s these relationships that help Lotame Semantic Classification to understand the context of the article.

What Happens After the Keywords are Collected?

This is where the magic happens. After the most important keywords, with the correct context and meaning, are identified, they are organized in Lotame’s standard behavioral taxonomy. We have taken thousands of keywords spanning topics from entertainment, politics, education, business, current events, and more and mapped them in our DMP for easy audience creation and ad targeting. Importantly, Lotame continues to build our keyword pool, add new relationships, and make connections to our standard IAB-compliant behavioral taxonomy.

What are the benefits of Semantic Classification?

  • Redefine your first-party data; find additional data points not currently available to you.
    • Drill much deeper to uncover previously unknown data points
    • 1000s of categories drive higher granularity (and higher CPMs!)
  • Improve audience targeting accuracy.
  • Stop relying on editors to identify concepts of an article, let a machine do it automatically – and with consistency!
  • Increased coverage, works in conjunction with metadata/breadcrumbs, or where metadata/breadcrumbs don’t exist, are inaccurate, or are insufficient.

While we’re still several years away from a Skynet situation, machines and artificial intelligence (AI) are here to stay and, for the time-being, here to help us improve many aspects of our digital universe. Want to see how machine learning is improving other aspects of Lotame’s technology? Check out what our Cross-Device and Audience Optimization solutions have to offer! Or, if you want to talk to someone about Lotame’s Semantic Classification solution, contact us

Request A Demo