Understanding Topic Modeling for SEO

You probably know your online content is essential for SEO. Matching your subject matter with what people are looking for online is crucial to drive the right traffic to your website. But you might not know SEO has become increasingly sophisticated, as Google has repeatedly introduced algorithm changes that return relevant search results. Topic modeling is a complex way that search engines determine that your content is what the searcher is really looking for, taking into account the other subjects you discuss in your text.

History Lesson: Keyword Density

In the early days of Google, keyword stuffing was common. Online content creators would analyze what specific phrases people keyed into search engines and repeated those words as frequently as possible on a page. While this initially resulted in strong SEO, with keyword-heavy sites ranking high on results pages, Google soon got wise to the tactic. In future updates, keywords alone no longer determined what sites got top billing.

Determining User Intent

It’s no secret that most people use everyday language to search for topics of interest. Search engines evolved to look beyond keywords to other hints that a particular website was an appropriate result for a search query. The starting point was synonyms of commonly used words. For example, it you typed “financial consultant” into Google, you would also get results for “financial adviser.”

But often, search queries are more obscure than that. You might want Google to refresh your memory by getting it to answer a question lingering in your head. For example, if you type in “that actress who believes in reincarnation,” you may get a long list of results about Shirley MacLaine. This evolution of search engine intelligence is based on a ranking system that connects certain words to ideas in your content.

This is sometimes called Latent Dirichlet Allocation, or LDA, which is a kind of natural language processing. Certain words are connected to other concepts or ideas, both in the real world and online. Increasingly, Google is creating search functions that respond to how people think and conceive of search concepts.

Hummingbird Update: Topic Modeling

For any search term, there are thousands, frequently millions, of results. They are a collection of every site indexed by Google that in some way references or discusses the search term. The problem is that each Google user is looking for something a bit different when they type in a search term. In order to rank results to offer the greatest relevance to the user, Google plays a bit of a guessing game, assuming that the user is like the most number of people and wants websites that cover the topic in the way that interests the greatest number of people. This was rolled out in full after Google’s Hummingbird update in 2013. In addition to words, Hummingbird is about context.

Sound confusing? Think again of Shirley MacLaine. As a Hollywood legend, she has a lengthy filmography that’s spanned several decades. But in recent years, she’s also become known as an author interested in paranormal topics. She’s also the sister of another actor, Warren Beatty, which makes her family another potential area of interest. In the absence of more specific information, Google will rank the sites about Shirley MacLaine according to whatever aspect of her life seems to interest the most people.

Topic modeling is a kind of ranking system, which decides what context of a particular search term means the most. Using fake numbers, let’s assume that in the case of “Shirley MacLaine,” interest is highest for her movie career and scores a 10, her books a 5 and her family a 2. Sites with content that discuss Shirley MacLaine and movies will rank higher than those that talk about other topics.

How are those scores generated by Hummingbird? According to Moz, it’s in part by looking at all the documents on the web that discuss a particular topic. Specifically, the algorithm connects the search term with related ideas. In order to determine ranking, the search engine measures how far away a key term is from a contextual determinant. For example, the word “movies” may be very close to the term “Shirley MacLaine” on a preponderance of documents, but far away from “Warren Beatty” on others.

Optimizing Your Page Using Topic Modeling

With the introduction of Hummingbird, strong content became infinitely more important than keyword density. Content creators now focus on a few key elements in order to get Google to connect them to the right searchers. First, they use synonyms in order to reach everyone searching for the topic, while still producing engaging copy. Second, they write well-organized content that provides Google with topic context that reaches the segment of the population that is interested in whatever their website is about.

You can use the services of a content marketing agency to optimize your SEO from a topic modeling standpoint. There are also several online models you can use for tips to structure your content, including your page titles, subheadings and text, to create the strongest online presence for your website.

Is it all too much? Contact our agency to learn more about what we can do to help. We’ve got SEO in the blood and love to get in-depth when talking about Google algorithms. Speak to us today about how you can optimize your site’s topic modeling.

Source List:

https://moz.com/blog/topic-modeling-semantic-connectivity-whiteboard-friday

https://msdn.microsoft.com/en-us/library/mt762914.aspx

https://www.mirumagency.com/blog/topic-modeling-semantic-search-seo-always-evolving

https://moz.com/blog/hummingbird-unleashed