Posted to Leigh Carver's blog on October 14th, 2013

What Every Marketer Needs to Know: Semantic Search Isn’t Smart.

Can a search engine answer a simple question? First WolframAlpha tried it, and now Google has hopped on board, but clearing up the confusion around semantic search isn't simple.

Semantic search is on the rise, and will be changing the face of the marketing and Artificial Intelligence (AI) landscape in the next few years. But to take full advantage of the benefits of semantic search, marketers need to understand its limitations, too.


What is Semantic Search?

Semantic search, or searching using natural language, has been increasing in popularity over the past few years. From typing in long-tail search queries to asking your phone a question ("Siri, what is semantic search?"), the search landscape has been in ceaseless evolution since the Archie search engine premiered in 1990.

Up until recently, smartphone-enabled (or spoken) semantic search has only been available to those with access to the WolframAlpha-powered Siri. But with its recent Hummingbird update, Google stepped up to the challenge laid down by Apple and entered the fray.

According to an article from Forbes, Scott Huffman, an engineering director at Google, claims that the search giant wants its algorithms to get to the point where people are able to have "a natural conversation" with the search engine. Hummingbird is the first step in a long journey towards this eventual goal, which-- given the technology we currently have-- won't be attained any time soon.

Here's why.

Computers Can't Do Natural Language Processing

While our current search and language processing algorithms are passable, the probability of being able to carry on "a natural conversation" with a search engine is low. In spite of the fact that great strides have been made in Natural Language Processing (NLP), a machine (or cluster of machines) attaining human-level NLP any time soon is highly improbable given our society's current technological and algorithmic standings.

In technical terms, the problem with bringing machines to human-level NLP is an AI-complete issue. A computer's fundamental lack of cognitive capabilities means parsing and identifying contextual clues and social nuances is an as-yet unsolved problem. NLP requires bringing computers to actual human-level intelligence-- and we don't have the technology to do that… Yet. What this means, then, is that AI algorithmically apes human behavior rather than organically replicating it.

In layman's terms, the problem with bringing machines to human-level NLP goes something like this. As good at computers are at playing chess, give a machine a simple question (a.k.a. Winograd Schemas) to parse and the algorithm breaks down. One of those schemas might look like this:

The large ball crashed right through the table because it was made of Styrofoam. What was made of Styrofoam?

a) The large ball

b) The table

Even our most sophisticated AI is terrible at answering that question, or other similar contextual schemas. And, believe it or not, NLP capabilities are really important for marketers to know about as we move into a semantic search-driven future.

Content Marketing, NLP, and Semantic Search

Google can parse your content, identify keywords, and try to match your website with a user's query, but as a machine-driven intelligence, Google is fundamentally incapable of understanding (or perceiving) the intended meaning or significance of your website. So as a marketer trying to create and promote content that Google will correctly match with user-intent semantic queries, you need to do several things:

  1. Be clear in your writing and optimization.
  2. Focus on generating content that matches up well with user-intent.
  3. Use every (SEO friendly) tool and module you can to send signals to Google about your content.

As we move into an era of "shinier" and "more sophisticated" search, it bears remembering that you should always be clear. Write it on a notecard and tape it to your monitor. When you are writing on your website, you should be writing for your audience… and you should be writing for Google. This means refraining from keyword stuffing or using any spammy SEO tactics.

Focus on clearly articulating the purpose and content of each page on your website in the metadata. In the body of the page, use lots of straightforward, keyword-rich H1 and H2 text. Build your content around user intent, not around keywords, and employ A/B testing to increase conversions and page stickiness. And last but not least, use Google-friendly tools (such as including rich snippets with the Drupal module) to help the search engine better understand your website.

If you're experiencing any doubt on whether your website is clear enough for Google to match up your products, services, or blog pages with semantic queries, it may be time for you to revisit and simplify your content. In content marketing, less is often more-- so be clear about your content. Use keywords in clearly-articulated sentences to signal to Google (and other search engines) the validity of your content.

One of the things I like to do is include sub-headers in my content as often as possible-- and, when it's natural, include keywords (such as Semantic Search) in those headers. The headers signal to users the path of the narrative the content will take-- and it allows Google to understand how the content meets long tail, semantic, and short tail queries. This, combined with optimizing the metadata for each piece of content, helps Google deliver my content to users who are most likely to benefit from it.

Drupal: The Semantic Search Marketer's CMS

Drupal in particular makes optimizing for semantic search easy. The CMS’s high level of customizability makes it a great framework for content marketers to work within. Extensive, SEO-friendly editing capabilities make it easy for marketers to communicate with Google about the content, while the extreme flexibility of content types makes it easy to design user-friendly pages that contain, highlight, and emphasize all the most important information.

Furthermore, the huge span of modules to choose from make it easy to build search-friendly content:

  • The Rules module allows marketers to build dynamic content that intelligently matches pages with user intent.

  • The module allows you to write tags that communicate to all of the search engines about the content of your page.

  • The Meta Tags and Page Title module allows you to communicate specifically with Google about your page.

  • The CKEditor allows marketers to easily create richly formatted content without having to be a programmer.

If you're looking for more information, my colleagues have generated some great content on SEO, user intent, and signaling information to Google that I recommend checking out. Mitch Holt put together an excellent video on user intent in organic search, Katie Thomas has an amazing article on using rich snippets to communicate with Google, and Chris Gaffney has some great advice on fresh content, Google, and your users.

Have a comment? Sign in: