NLP in SEO: How Natural Language Processing Affects SEO

NLP in SEO: How Natural Language Processing Affects SEO

I think that by now, we’ve all come to the realisation that SEO is not a ‘one-time thing’ but rather an ongoing process.

The rules of the game are continuously changing and evolving, and you can’t always guess what’s coming next.

In other words, what works today won’t work tomorrow.

But for today, I’m focusing on one of the mega-developments that came about and dominated the scene ever since; Natural Language Processing, or NLP, as we know it.

If you still haven’t gotten the gist of why it’s so important to search engines, keep reading this blog post to find out exactly what it is, how it relates to SEO, and how you can use it to improve your content marketing efforts.

What is NLP and its Uses?

Just like its name implies, natural language refers to the language that humans speak in their everyday lives, regardless of what that language is.

Then comes the processing part; primarily, NLP exists in an attempt to program machines to comprehend queries said or written in natural language. And put a line under comprehend, because that entails understanding the intent behind what’s being said, too.

All of us are exposed to NLP’s technology every single day, but it’s become so natural that we rarely realise it.

Let me give you a small example: have you ever wondered how your Gmail emails are auto-completed? Yes, that’s NLP! How about predictive text? That, too!

Even on Facebook, did you ever search for the name of a page that you’re absolutely certain isn’t right, only to get the exact result you were looking for? That again, my friend, is NLP in action.

But then again, what does that mean for SEO? Let me tell you.

In October 2019, Google announced that it’s introducing:

a significant improvement to how we understand queries, representing the biggest leap forward in the past five years, and one of the biggest leaps forward in the history of search.

The update, coined BERT (Bidirectional Encoder Representations from Transformers), predominantly relied on NLP to “consider the full context of a word by analysing what comes before and after it,” and accordingly, understand the user intent behind search queries.

By applying the BERT model to both rankings and featured snippets on Google’s search engine, the search results displayed are guaranteed to be relevant and useful to the reader. Now, users can make queries using 100% natural language and still receive quality content.

In essence, since Google’s algorithm uses NLP, both on-page and off-page SEO are impacted. Why so?

Because syntax analysis is involved, carefully analysing content for both its meaning and context, along with whether it’s well-written and purposeful. That makes a lot of SEO tactics infeasible, especially black-hat ones.

Therefore, what was once an explicit keyword-based search is no longer that. Rather, it’s a search based on context and intent, which isn’t a minor change.

As software engineer Amit Singhal puts it:

Fundamentally, it’s the difference between ‘give me what I said’ and ‘give me what I want.’

How Exactly Does NLP Enrich Search Quality?

According to Google, 15% of search terms being used today are being used for the first time ever.

The common pattern we’re seeing is that long-tail queries are back again, especially with the availability of voice search.

That actually creates a challenge for Google.

You and I know that historical data plays a massive role in successfully delivering relevant results for the user, so what if there’s not enough historical data?

Using NLP is the answer here. The search term at hand can be analysed by natural language processing, and voila, user satisfaction!

Where Do I Start?

By now, you should have a clear idea of what NLP is and what it does, but you’re probably still looking for something tangible to give you a head-start.

Been there, done that. This is where I’d like to introduce you to Surfer’s NLP Analysis feature.

Introduced in January 2020, the feature includes Sentiment Analysis, Entities Coverage, and Context Usage, each with its own purpose.

In details:

  • Sentiment Analysis enables you to find out whether the context of the top-ranking pages is positive, neutral, or negative
  • Entities Coverages gives you better suggestions with regards to entities that are most relevant for Google for the given topic and
  • Usage Context gives you data on how your competitors are using relevant words and phrases

Let me explain these, one by one and tell you how they relate to your content and NLP.

But before I do, I’d also like to share with you Google’s Natural Language API demo, which is quite an impressive tool that enables you to check text for free.

Google Natural Language API demo
Google Natural Language API demo

Using Google machine learning, you’ll be provided with insights about the unstructured text of your desire, including the same ones mentioned above on Surfer.

Back to sentiment and entity!

Sentiment in NLP

Sentiment in NLP, simply put, means emotion. With regards to content on your website or in an article, it refers to the undertone conveyed through the content.

The algorithm Google uses measures sentiment value for both the entire subpage of a website and the subsections of the content.

When content sentiment is positive, that indicates that the topic is discussed favourably. Such examples include commonly used positive keywords, including “outstanding”, “astonishing,” “champion,” and the like.

To spot positive sentiment, look for values between 0.25 and 1.0, with 1 being the most positive.

By the same token, negative sentiment refers to the opposite. In content with negative sentiment, you’ll encounter tons of words such as “horrible”, “inferior”, “poor”, “hazard”, and more.

Values of negative sentiment range between -1.0 and -0.25.

Last but not least, neutral sentiment is when content includes a mix of both positive and negative sentiments. This is often common in unbiased product reviews that cover both sides of the spectrum.

In this case, the range is between -0.25 and 0.25.

Did you stop for a second to think about what the purpose of this aspect is?

It’s to push content that’s created for user satisfaction, rather than for the search engine.

Connect with your audience, convey emotion, and you’ll rank with NLP!

Entity in NLP

Entities generally refer to named objects that can be identified and classified. Famous people, numbers, organisations, events, products, locations, and more are all examples of entities.

What NLP does is that it selects and evaluates the common entities found in your content.

With Google categorising and classifying entities, the search engine is better able to satisfy the user intent and present relevant results.

Relating to this point is also salience. In NLP, salience refers to the entity’s importance in the content.

Ranging from 0.0 to 1.0, higher salience values indicate more entity importance and relevance for the content. For instance, “sunset” may be more important and relevant than “sunrise” when we discuss the beach.

That fact also has to do with categorisation, which refers to Google’s ability to distinguish between the different content topics being discussed and classify them as needed.

An article about software would be in a different category than an article about art, for example.

Is This the End of the Keyword Research Era? 

Not necessarily.

However, there’s no denying that BERT and NLP do raise a red flag when it comes to solely depending on the traditional keyword-based search for SEO.

If you think of it, we don’t always search using structured data or well-written sentences.

And as a matter of fact, a search engine like Google doesn’t really need keywords to understand the nature of the content. Natural language is quietly taking the seat instead.

More often than not, we go to Google in a rush and proceed to write whatever crosses our mind, the way we would say it out loud.

Things like prepositions, question marks, and more, are very often missing in our queries. We do search using natural language, and that’s NLP’s role; to make sure that the context and sentiment of each word are understood.

Take a look at the following example by Google.

Previously, some queries were just too sophisticated to be assessed, like this one.

Before NLP, “math practice books” were the keywords that were picked up on according to machine learning. In contrast, now, the intent came into play as well, also recognising the importance of prepositions.

But let me just make this clear: NLP is not a threat. In fact, this can be an easy road to take.

If every piece of content you develop is to-the-point and well-structured, you really have nothing to worry about. You should only worry if your web pages are keyword-stuffed with no effort to better understand your target audience.

Bottom line:

all you need to do is write using natural language, focus on context, and put less and less focus on keyword density.

Actionable Tips: Optimise Your Website for NLP

I know that I’ve been stressing on the fact that there isn’t really anything you can do to optimise your website for NLP.

And you’ll find that I’m not the only one saying so, but is there really nothing to do?

The best practice will be to use natural language, but there are some tips that can make a difference as well. Let me walk you through them.

Use Google Search Console

Understandably, after the BERT update and incorporating NLP changes, there were very noticeable changes in SERPs, now that Google understands almost every single query in a better way.

It’s important to invest some time in doing keyword research to pinpoint the query types that witnessed a traffic increase, along with those that suffered drops in organic traffic. Make particular note of these to ensure content relevancy in each piece of content you produce.

It’s also a wise move to analyse your situation in comparison to your competitors, especially when you spot a keyword that dropped in rankings. Compare the content to identify issues that may have caused the drop.

To pinpoint these keywords, do the following:

  • In Google Search Console, click date, filter, and compare October 2019 to November 2019
  • Choose “impressions”, and sort them by “difference” to see the queries that were hit hard by BERT
  • Switch between “queries” and “pages” to determine whether the issue is with one page or several ones

Structure Your Website

There has been talking going around from search algorithm experts that internal linking and website structure will play a role in the process of natural language processing and using NLP.

With that said, this is your queue to start working on optimising the structure of your website if you don’t already have an established one. Work your way top-bottom, from the whole domain down to each subpage.

For the website as a whole, that entails revisiting your internal linking, anchor text unification, breadcrumbs implementation (if applicable), and navigation.

Then, when it comes to the content and working article by article, start by eliminating all keyword stuffing. Following that, work on your headings, schema, sources, title tag for each page, topic coverage, entities, and sentiment.

Optimise Your Website for Mobile

This probably isn’t news for you, but over 60% of searches are done using mobile devices, and as an immediate result, search engines started endorsing mobile-friendly websites with relevant content SEO-wise over those that aren’t.

That’s been the case for around 4 years now, so why am I bringing this up again? One more time, voice search.

The virtual assistant marketing is becoming flooded, and so are voice commands!

Therefore, there’s basically no point of aiming to rank high if your content isn’t optimised for mobile.

And to make it clear, I’m not referring to the website design. In essence, what I’m referring to is structured data and content.

Make your content easy to digest, even on the smallest screens. Use larger fonts, shorter article paragraphs, clear subheadings, bulleted lists, and other formatting methods that make your content appealing.

Adjust Internal and External Backlinks

Natural language processing now empowers Google’s algorithm to assess the context and relevance of both internal and external links, which both play an undeniable role in SEO.

It goes without saying that when a link is found to be positioned in the right context, it gains a much higher value on search engines than if it were irrelevant.

That’s really a tip on a gold platter for you: work on your links to situate them in logically-related parts of your content. It’ll work wonders with your SEO!

Snippets, and More Snippets!

Featured snippets are the way to go moving forward! Make use of this and try to take over as many as you can.

“Frequently Asked Questions” are a reasonable starting point. Why?

Because voice search queries tend to be in the Q&A format, and that’s exactly what your FAQs will do.

Just a tip, make your questions very specific, just like a voice command search would be. Making the most common queries available on your website will give you a huge edge with search engines.

Focus on Long-tail Keywords

To fully make use of the SEO benefits associated with natural language processing, content should be focused on long-tail keywords to rank higher for voice search, in particular.

If you were to make a search query using your voice on Google, you’d probably use full sentences that are much longer than the usual keywords you would use if you were using a search engine on your laptop, right?

That makes sense because sentences are more conversational.

That’s why, with the increasing popularity of voice, your content needs to provide a diverse range of keywords, and pay more attention to long-tail ones to not miss out on the new SEO opportunities.

Not to mention, long-tail keywords are already less competitive, and therefore, easier to rank for!

For example, instead of using “SEO specialist in Essex” as your primary keyword, you would use “Where can I find an SEO specialist in Essex?.” Small difference, but better results!

Make what, how to, how, when, who, where, and why your new best friends, since those are the most common among voice queries.

A Trial Example: The New SEO Process With NLP

There are tons of case studies out there about using natural language processing, but I’d still like to walk you through the process itself behind the scenes to help you with the new SEO process.

I’ve already acquainted you with Google’s NLP API, which will provide you with the sentiment, entities, category, and salience score for the text you provided.

Now, let’s start our competitors’ analysis to know where we stand.

1. Identify Your Main Keyword

What’s the keyword you want to rank for? Do your keyword research and choose accordingly, with NLP in mind.

2. Pinpoint Your Competitors

Don’t fall into the trap of thinking that your competitors are those ranking higher than you in search engines. It doesn’t work that way.

Instead, you actually need to specify your own area and those working in it.

Start by determining your content type: do you focus on blog posts, articles, videos, landing pages, infographics, or others? Any website that doesn’t offer that same content is to be excluded right away.

Then, also omit any outliers when it comes to word count; websites that tend to go way above or way below the average article length.

Lastly, websites that are merely ranking because of authority shouldn’t be among your competitors, especially now, given NLP.

3. Record Your Data

I can’t stress this one enough. To best make use of your data, you should compile it for easy reference.

Personally, I’m a fan of this template by Surfer for NLP competitors’ analysis.

It’s ready to use and even has the fields jotted down. Give it a shot and watch the effects of structured data!


4. Compare Your Content Against Competitors’

Make sure to “go with the flow,” especially when it comes to sentiment. It’s not a good call to go against the preferred sentiment.

If you think it’ll make you stand out, it’ll actually take your SEO down big time.

For example, if all the top search results are reviewing the Apple smartwatch positively, and you decide to opt for a negative review, don’t expect positive SEO trends.

Simply, Google has already compiled the data over the years about this particular topic from the top ranking websites and based on sentiment analysis, positive reviews are the most dominant.

Therefore, it’s a good idea to set a benchmark and compare yourself with your direct competitors to better understand user’s search queries in order to tailor the content to their needs, and hence, achieve higher SEO rankings.

5. Implement the Changes!

Now that you have gathered all the needed data, it’s time to test NLP!

Make the necessary tweaks that have become visible through your NLP audit using either Google’s NLP API or Surfer’s NLP Analysis feature.

If you’re lucky, you’ll actually find that not all your content will require too much updating, especially if using natural language is your default!

However, just a few changes do go a long way both with NLP and SEO.

NLP in SEO: Wrapping Up

Machine learning and its ability to use NLP is truly fascinating, yet it’s a world of its own.

In this article, I just scratched the surface to make things less complex while providing you with the essence of what the buzz is all about.

When all is said and done, the key to take your SEO higher is simply using natural language in your content.

In a nutshell; that’s how natural language processing, NLP, works.

Also, we’re lucky to have so many tools available to lead our SEO processes.

Just be sure to utilise them fully and always have your eyes wide open to any changes in SEO practices to stay up to date with the industry and its requirements.

And to conclude, if you need support with your SEO audit or NLP analysis, don’t hesitate to reach out to me!

You might be interested in …