It’s a fact, the Google BERT update that arrived at the end of 2019 left no one indifferent. But who is this new algorithm equipped with artificial intelligence? Does it really represent the future of natural referencing? SEO players, wary at the start of deployment, are now unanimous. BERT has succeeded in breaking through the expectations of Internet users and promises to be the start of a search engine revolution. Do you want to understand this robot of the future and see what it will bring to your website and search habits in the long term? Here I give you some keys.
Who is Google BERT?
In order to better understand this machine learning algorithm, let’s decipher its evolution, as well as how it works.
The birth of the BERT algorithm
The story of BERT, by its full name “Bidirectional Encoder Representations from Transformers ” unofficially begins in 2018. It then left Google’s laboratories to be released as open source. Scientists, specialists in artificial intelligence and natural language processing, therefore freely appropriate this new system of linguistic approach and enrich it. Derivatives of BERT also arise as:
- the Roberta algorithm created by Facebook;
- the MT-DNN designed by Microsoft;
- or our patriotic Camembert, a French version developed by the National Institute for Research in Computer Science and Automation (INRIA).
Google is very influential in the field of artificial intelligence and its fields of study, including machine learning. It has already used AI in its previous updates, such as Rank Brain. With BERT, however, he enters the air of artificial consciousness. This algorithm is more efficient than its predecessors in terms of results and teaching speed.
The Google BERT Mechanism
Bert is a small specimen that combines the natural language encoding methods of the “Transformers” and the data processing faculties of the Clouds TPU.
Here is an example of what this means in practice.
As you read this article, you focus on each term. Your mind, at the same time, will remember the essential keywords in order to grasp the overall context of the text. The work of Transformers is therefore to analyze the words of a complex request, to put them in relation in order to grasp the semantics of the sentence and to better understand its global meaning. Clouds TPU are integrated circuits that accelerate the workload of Transformers in order to make them faster and more efficient.
This algorithm is, as its name suggests, bidirectional unlike its predecessors which were unidirectional. It therefore reads our requests typed in the search engine in both directions to allow us to better understand our requests. I’ll give you an example to better understand:
Before, algorithms could only read the sentence one way to try to understand the context. Moreover, they studied the terms one by one. For them, it is impossible to grasp the exact meaning of the sentence.
BERT uses bidirectional language modeling. In addition to analyzing the words in relation to all the others, it reads the sentence from side to side of a target word in order to place it in context. BERT understands here that the word “pear” is not the fruit, but is used in an expression that indicates “to share something”.
We are no longer dealing with an indexing engine, but with a robot that learns and progresses in its analytical capacity. It is also becoming popular in the closed world of natural language specialists. We are slowly starting to talk about the phenomenon of BERT ology. This new research on natural language processing techniques (NLP) has, on the other hand, led to remarkable improvements in the fields of automatic translation or among personal assistants such as OK Google, Siri, Cortana or even Alexa.
The consequences of BERT on natural referencing
In order to allow websites to stay on the course of the 1st search page following this update, I offer you some tips to adopt.
Impact for SEO
The Google BERT update is designed to better capture and understand voice searches and complex user requests. The changes made by the update, however, only affect one in ten requests. The impact of BERT is therefore low for websites. Google told SEO specialists that they could not optimize their website for the algorithm. However, there are techniques to follow in order to please our new friend.
Best practices for the future
In the continuity of its updates and with the rise of voice queries on the web, the algorithm favors sites that work on the search intent of readers. To allow you to improve the visibility of your site in the future, to satisfy Internet users and the algorithm in your next articles, you can now adopt good habits in your editorial strategy:
- present texts with quality content;
- use long-tail keywords to offer precise and detailed queries and capture more organic traffic;
- use stop words (le, la, du, etc.) which, with BERT, become very important in overall readability;
- integrate the semantic cocoon into your articles in order to respond precisely to the needs of Internet users and improve your visibility.
If you lack the time or experience to write your articles, do not hesitate to call on the Ghostwriting Agency of the Web, whose job is to design content in your image. The BERT algorithm, a nod to the Muppet Show character, is quite different from previous updates presented by Google. Indeed, resulting from artificial intelligence, it evolves and progresses with the Internet user. If you apply yourself to satisfying the curiosity of your readers and to creating qualitative content with a rich lexical field, the Google algorithm will not hesitate long to propel you to the 1st page. If you wish, you can now tell us about your project.
Google BERT update and its impact on user experience
Considering the possibilities offered by BERT, let’s see concretely what it brings to Internet users during their searches on the web.
Actions of the algorithm on the searches of Internet users
BERT officially arrived at the end of October 2019 on the web for English queries. Then, at the beginning of December 2019, we discover it in more than 70 other languages, including French. Google’s goal is to develop an algorithm that has a better understanding of text and voice searches on the internet. Indeed, it offers its user more targeted responses. The BERT algorithm is also able to perform several actions:
- understand ambiguous phrases or sentences, as well as the contextual meaning of terms;
- analyze linking words, pronouns;
- identify homonyms;
- automatically generate featured snippets;
- answer questions directly in the SERP;
- adapt even more to the growing rise in demands made on the voice assistant;
- understand long and complicated questions and predict your next sentence.
How the Google BERT update works
Before, search engines focused on keywords. Now the algorithm can read between the lines. It will analyze a sequence of words in order to understand the meaning of the sentence in its context. BERT is able to decipher our feelings and decode the subtleties of each language. The UX (the user experience) will therefore be more adapted and sharper. Thus, BERT, one of the 3000 Google algorithm updates of 2019, is unquestionably the most important. The multinational tells us that it would even be “the greatest advance in 5 years in the understanding of research”.
