5 keys to understanding BERT, Google's new algorithm
BERT is Google's new algorithm optimization, which has been developed to improve users' search results.
Last September Google announced the launch of the biggest update of its algorithm in the last five years. BERT ('Bidirectional Encoder Representations from Transformers') then began to affect English searches, however, from the last December 9th it is now also operational for searches in over 70 different languages and more than 20 countries.
What is Google BERT?
BERT is an open-source neural network based on natural language processing, which aims to understand users' searches as accurately as possible. The breakthrough lies in its ability to interpret human natural language, including context and nuance. Until now, the algorithm interpreted searches word by word, but thanks to this update Google is also able to understand the connections between the words in a search.
How does bert work?
The machine learning model used by BERT is based on an idea put forward by the British linguist John Rupert Firth in the 1960s, in which he stated that it is possible to deduce the meaning of a word from those around it.
BERT uses artificial intelligence to enable Google to interpret the terms that make up the searches that users type in both directions. In this way, it can deduce what each word in a search means concerning its context and what connection it has with the other words that make up the search. For example, if a user types "change laptop brightness", they are probably looking for directions on how to adjust the brightness of their laptop screen. The word "change", therefore, in this case does not mean "replace one thing with another", but rather refers to the meaning of "adjust".
What led Google to develop this update?
Google's main goal is to provide the best results to users as quickly as possible by providing them with accurate, quality information and redirecting them to trusted websites.
Pandu Nayak, Vice President of Google Search, says that about 15 percent of the searches performed daily on Google are new, not previously performed. For this reason, the analysis of terms by coincidence that Google's algorithm has been carrying out until now was not sufficient to provide accurate results for new searches.
Another motivation for BERT's development has been the significant increase in voice searches, especially in countries such as the USA, which has made it necessary for Google to be able to adapt as accurately as possible to natural language. Therefore, BERT will also affect those users who request information from Google Assistant by voice.
What has the arrival of BERT achieved?
To understand the great leap forward that Google has made with the arrival of BERT we can see another example that clearly shows the progress made:
This year 2019 a user did the following search: "2019 travel from Brazil to the United States requires a visa". Among the results that Google returned was an article from the Washington Post about traveling from the United States to Brazil. What happened, in this case, was that Google was not able to assimilate the importance of the preposition "a", and the results it offered did not correspond to the user's need. With the arrival of BERT, in this same search, Google offers as the first result the page on visas for tourists from Brazil from the official website of the U.S. Embassy in Brazil.
What BERT has been able to correct is the mismatch that occurred in some searches between the user's search intention, what he expressed in the search itself and what Google understood.
How will it affect search engine positioning?
With BERT, Google positions in the first places to search those websites that care about responding naturally to users' queries through the content they offer. In this way, we must start to consider SEO as a set of techniques to optimize results by focusing on users, and not so much on the traditional Google robots.
The improvements in the understanding of natural language achieved by Google BERT and its efforts to better understand queries will have a positive impact on those websites that offer content written more naturally. In this way, the information on each website is understood by Google in a much more natural way, so in the future, everything points to the fact that SEO techniques that generate "unnatural" texts in terms of language use and layout will be less and less necessary.
Because of this, it is not possible to optimize a website specifically for BERT, since it is not a positioning factor. Its function is to send higher-quality traffic to our website when the information meets the user's needs, even if, for example, the search engine keyword is not explicitly present in our content. It is a more "human" algorithm, which seeks to understand the natural language of users.
From Protecmedia we encourage you to celebrate this type of change, which generates so much uncertainty, as Google is constantly working to optimally connect users' queries with the information that is most accurate and useful. BERT is a positive development for organic positioning, as it allows us to write content that is more conversational and natural. In this way, we will not be so conditioned by the static use of 'keywords', which sometimes make texts lose their naturalness.