Google DeepRank: The Algorithm That Made Search Natural

By Emma Grant  

| December 30th 2020

Google is an incredibly clever thing, and it would seem it’s getting more and more intelligent by the day. The good news for internet users is that it largely focuses its intelligence on improving the search experience. So every time you hear word of an algorithm change that you know your SEO agency will be factoring in to your search marketing campaign, you can rest assured that as an internet user, it will have got even easier for you to find precisely what you are looking for online.

One very important algorithm which delivers more relevant search results by understanding language in exactly the same way we do as humans was introduced to us in 2019. Its name was BERT.

Now, we looked at BERT when it first surfaced, and we were suitably impressed at how it had become the superpower behind search intent.

Knowledge is power (and search intent is knowledge)

Regular readers of the Figment blog will know we go a lot on the importance of search intent. It’s a crucial tool in many of our SEO and content marketing strategies.

If you can tell by the very nature of a search phrase whether the user is looking for information, trying to find a particular website, seeking to compare products ahead of settling on a purchase, or actually ready to make a purchase, this is very powerful knowledge.

Because with this knowledge, you can optimise the right web pages to suit the right intent, which will in turn help to boost your click through and conversion rates. And what do boosted click-throughs and conversions lead to? More sales!

So, back to BERT. This friendly sounding algorithm has the job of analysing the search intent and in-context meaning of search phrases, ahead of returning the most relevant results.

BERT stands for Bidirectional Encoder Representations from Transformers. No wonder they call him BERT! BERT has the ability to understand search intent, cross-reference the context, and analyse relationships between words.

You know when you’ve been actively researching something online, you go to type yet another query into Google, and it automatically seems to know what you’re looking for, suggesting searches that are totally relevant? This is BERT, associating your previous searches with your current search. Clever stuff. And what a great triumph for user experience.

Introducing DeepRank

Just recently, Google released a video about how search works. The video that has grasped the interest of many a SEO agency in London revealed new details about the making of BERT, which has now evolved into the DeepRank algorithm.

DeepRank got its name from the deep learning methods used by BERT. Consider it the integration of BERT into Google Search, building more on the search engine’s machine learning and natural language processing capabilities.

Natural language processing (NLP) means you can ask a question of Google in exactly the same colloquial way you’d ask a friend. This is nothing new though. Google has actually been working on natural language processing for the best part of two decades. Spelling corrections came first, then along came understanding synonyms.

BERT has been doing a grand job of enhancing the understanding of natural language queries. Before his arrival, there was no real way for the search engines to get to grips with the nuances and subtleties of the everyday language we all naturally use, and the relationships between words and phrases.

The objective of DeepRank is to enhance the search experience, making it more intuitive, almost as if you are asking a fellow human for answers. How it’s got to where it is makes for quite an interesting journey.

Natural language processing and the importance of ‘stop’ words

Years ago, Google’s search algorithms would ignore the ‘stop’ words that were included in a search query. By way of an example, if you were to search:

“What time is the sun going to set tonight in London?”

All Google would have seen before natural language processing became a thing, was this:

“Time sun set tonight London”

But over time, Google started to realise that the stop words actually played an important role in communicating what people are trying to discover. It is these words that fill the gaps to reveal search intent, which means that the search results returned are way more accurate, and frustration-free.

With DeepRank, it is now possible to construct a search query naturally, rather than having to simplify or overthink your phrasing. But it’s not been an easy ride to get DeepRank to where it is today.

“How search works”, by Google

Months of testing and meticulous analysing of individual search queries were undertaken to ensure DeepRank was helping rather than hindering search results, and it’s fascinating to see some of this behind-the-scenes work in Google’s video.

It really does reveal the in-depth work that goes into an algorithm update, and it’s well worth a watch if you’re interested in the levels Google is prepared to go to in order to improve the user experience.

Watch Google’s video:

Want to increase your search engine visibility and online conversion rate in 2021? Talk to Figment, the SEO agency London businesses trust to increase their sales.

Here at Figment, we fully appreciate that when there’s Google algorithm update, it spells good news for users and their search experience. We also know that any update could mean a change in SEO strategy is needed. It’s precisely why we keep on top of the game in every respect, responding to changes as they happen, and preparing for future shifts.

If you are looking to partner with a new SEO agency in London or Surrey for 2021, why not talk to Figment? We’d love to show you how we’ve achieved exceptional results for our clients, results that have seen goals not just met but exceeded, and in super-fast time too.

To learn how we can make a difference to your sales and help you grow your business in 2021, please do get in touch.