Have you ever heard of Google BERT?
It’s the latest update to Google’s algorithm, based on an artificial neural network technique that allows the search engine to better understand natural language (NLP) as users express their search intent in queries.
This update, initially launched in English in the USA on December 9, 2019, has been rolled out in over 70 languages, with considerably fewer effects than expected by many browser users.
In this article, we will look at:
- what and why Google BERT was launched
- how Google’s algorithm update works
- examples of SERPs before and after the introduction of BERT
- the expected and actual impact of this update.
Want to learn more about how the BERT algorithm works to optimize your website’s ranking on the SERP? What are you waiting for?! Discover Digital Coach® SEO course.
Google BERT: what is it and why was it introduced
Google made a formal announcement on October 25, 2019, via its blog, declaring the arrival of Google BERT. They described it as the most significant advancement in the past five years and a major landmark in the history of Search.
But what is Google BERT, and what is the reason behind its launch by the well-known American company? Let’s find out in this paragraph!
A definition of Google BERT
BERT, short for “Bidirectional Encoder Representations from Transformers,” is nothing but the latest update to Google’s algorithm, based on technology that allows the browser to better understand the natural language (NLP) of users.
Nowadays, almost all online searches are made using commonly used expressions in conversation rather than just keywords as in the past.
Queries are becoming longer and more conversational, and prepositions within them play a crucial role because, depending on their interpretation, they can substantially change the meaning of the sentence.
Therefore, it becomes essential for the search engine to capture the search intent as precisely as possible. This allows for better user satisfaction and provides them with relevant answers, meeting their needs.
Specifically, the American tech giant predicts that the implementation of Google BERT will influence approximately 10 percent of the search engine results it produces. This change is expected to significantly improve users’ ability to locate their desired content.
What does Google say about the BERT algorithm?
The well-known company has explained that the reason behind the introduction of Google BERT was the need to make online searches simpler and faster, hence more effective.0
To achieve this, it was necessary to implement features that allowed the algorithm to understand the natural language used by users to conduct web searches.
In practical terms, this meant teaching the algorithm how to relate each word to the others in the same sentence to grasp the context. Here’s what Google stated on its blog at the time of the launch:
This significant advancement emerged from Google’s exploration of transformers, which are models designed to analyze words within a sentence by considering their relationships with all other words, as opposed to a sequential one-by-one approach.
BERT models enable a comprehensive understanding of a word’s context by examining the words preceding and following it, making them especially valuable for discerning the intent behind search queries.
Searches made with Google Assistant, the voice assistant that provides answers through Android or devices using Artificial Intelligence, will also become more effective thanks to BERT.
The SEO Specialist needs to have a good understanding of the Google ecosystem. Take the Test and find out if you are naturally inclined to this profession
How does Google BERT work?
After the launch of the BERT algorithm, many have limited themselves to investigating what it was and why it was introduced. Few, however, have sought to delve into how this algorithm update works, which is what we will try to clarify in this paragraph.
The bidirectional model of the new algorithm
Google BERT uses Transformer, a computer model based on an artificial neural network that imitates the functioning of human brain neurons. The characteristic of this system lies in including a “bidirectional” encoding mechanism that reads the sequence of words simultaneously.
Upon closer examination, the adjective “bidirectional” is not entirely correct; it would be more appropriate to say that this encoder has no direction at all. In this way, it would be more intuitive to differentiate it from previous models called “directional,” as they read strings sequentially, from left to right or vice versa.
This led the algorithm to interpret each word individually, risking not capturing the context in which it was used, and thus its real meaning.
Google BERT addresses this potential issue by simultaneously processing each term, enabling it to understand the overall context being discussed and the sense the user intended for each word and preposition.
BERT’s alignment strategies
Google’s progress in the field of Machine Learning and numerous testing sessions have allowed the latest update of the algorithm to employ two alignment strategies, aiming to connect the words within the same sentence.
Let’s explore them in detail:
- Masked LM (MLM): This technique involves masking and replacing 15% of the words in a sequence with a Token called Mask. The goal is to train Google BERT to predict the missing terms based on the meaning of the unmasked words, thus relying on the entire context of the sentence.
- Next Sentence Prediction: In this model, pairs of sentences are input to the search engine, extracted from an original document. In only 50% of the pairs, the second string represents the next sentence in the initial text; in the remaining cases, it is a randomly chosen sentence from within the document. The system is trained to determine whether the second sentence in each pair corresponds to the immediate proposition following the analyzed text.
Google BERT: some examples
Now, let’s look at two examples provided by Google on its blog to explain how the SERP changes after the introduction of Google BERT:
- Example 1: Typing the query “2019 Brazil traveler to USA need a visa” before the algorithm update, the browser provided links useful for US citizens intending to travel to Brazil. However, the correct search intent is to understand if Brazilian travelers to the US need a visa. As you can see from the example below, the updated algorithm intervenes by grasping the exact search intent thanks to its ability to understand the real meaning of the term “to”.
Screenshot from Google’s blog https://blog.google/products/search/search-language-understanding-bert
- Example 2: When asking “Can you get medicine from someone’s pharmacy,” the old model responded generically, explaining how to pick up prescription medications. Now, the system correctly processes the meaning of the word “someone,” perfectly capturing the search intent. It responds by explaining how a person can obtain prescription medications on behalf of someone else.
Screenshot from Google’s blog https://blog.google/products/search/search-language-understanding-bert
The effects of Google BERT
How will the SERP change after the launch of Google BERT? This is the question that many asked before the update’s release. Continue reading to discover the impacts it has generated.
Insignificant effects?
All in all, we can say that Google BERT has fallen short of expectations regarding the effects it produced. Many thought it would trigger a revolution similar to that caused by the release of the Panda and Penguin algorithms. The reality was quite different.
No significant changes have been observed on the SERP; in fact, the response links have remained largely unchanged, apart from some physiological fluctuations.
Google BERT and SEO: what’s changing?
There have also been a few impacts on the SEO of the algorithm update. This update focuses on analyzing queries rather than individual web pages, leading to a new and different view of the role of the Copywriter SEO.
While this figure used to write for the search engine, considered a machine in every sense, it is now called upon to write directly for readers and to focus on content quality and originality.
Download the Free Mini Guide and Learn how to create SEO-focused content effectively
The more valuable, useful, and aligned with search intent the content produced is, the higher the likelihood that the content will be rewarded by the algorithm and placed among the top positions on the SERP.
Apart from these considerations, it’s worth noting that Google hasn’t provided much guidance on how its algorithm has changed.
It remains a mystery what the complex mechanism is that interprets user queries and manages the placement of links on results pages.
It’s certainly not a coincidence; it’s a strategy by the American giant to prevent website owners from using this information to their advantage and positioning themselves at the top of the SERP at the expense of their competitors.
This way, everyone is encouraged to offer quality content to outperform competitors, engaging in a level playing field battle from which only the most skilled and deserving can emerge victorious.
In this article, we’ve seen what Google BERT is, how it works, and what effects it has generated. As you’ve learned, the benefits of this Google algorithm change primarily concern users, who are destined to receive increasingly precise and relevant answers to their questions.
The invitation to SEO Copywriters is, therefore, to focus on content and produce more valuable texts than ever to ride the wave of the most significant advancement made by Google in the field of Search in recent years.
Are you passionate about digital or would you like to learn more? Request more information about our courses and programs
Hello there! I’m passionate about a variety of interests, including playing badminton, diving into video games, indulging in movies and series, and exploring new destinations while listening to my favorite tunes. I thrive on learning, constantly seeking out new knowledge and experiences. My professional journey started in a call center, but I’ve transitioned into freelancing, dedicating myself to continuous improvement and striving to be the best at what I do.
0 Comments