With the introduction of Google BERT, in fact, we are faced with a substantial modification of the algorithms proposed for understanding search queries.
What is Google BERT and why is it so important?
BERT stands for Transformers Bidirectional Encoder Representations and represents a system for automatically processing language in a natural way.
BERT applies a bidirectional approach that, compared to previous left-to-right and right-to-left approaches, allows you to understand words within a broader context.
This open source technology allows anyone to create their own question and answer system.
This algorithm update follows Hummingbird in 2013 and RankBrain in 2015 and is possibly even more important than them. Pandu Nayak defined it as the “biggest leap forward in the last 5 years” and this gives us an idea of the importance it has in its field.
The system uses an artificial neural network, which realistically imitates how neurons work and how they communicate with each other.
The purpose of this technology is to help Google understand people's increasingly correctly spelled information searches to deliver results relevant to their needs.
The models applied by BERT do not act on words taken as unique entities but considering the previous and subsequent ones: in this way it is much more plausible to understand the real search intention.
This introduction addresses the rise of voice search and devices like Google Home to provide more relevant answers to search needs.
This is the result of a long work carried out by Google over the years that required numerous tests and that was able to materialize thanks to processors that work by connecting words as groups and not as single entities.
During testing, operators removed about 15% of words from random groups of words and urged the machine learning mechanism to fill in the gaps, running millions of simulations to identify the right options to return full meaning to the sentences.
Google BERT acts on the words considering for each one the one that follows and the one that precedes and therefore has a more complete awareness of the context.
BERT does not act directly on the positioning of websites, but on Google's ability to understand the meaning of queries written. Bert is estimated to impact around 10% of English queries in the US, making it one of the largest algorithm changes in history.
At the moment it is not yet active in Italian but Google plans to expand it to several countries in the coming months.
A very noticeable advantage is that neural networks facilitate intuitive adaptation to other languages without the need for ad hoc adjustments.
To give an example, in a search like “2019 Brazil traveler to USA needs visa”, the algorithm would have first analyzed the query word by word, whereas now it can examine it as a whole.
Google BERT now also takes into account the reading order from left to right: while before all words had the same weight, now the system is able to capture the importance of words based on their position and also takes into account the prepositions.
In this way, Google meets the needs of understanding even informally written queries with a decent level of accuracy, as we are used to doing with voice devices.
This technology will also be applied to featured snippets, which will likely result in a change to the results served in featured boxes.
However, this is not a definitive change: the path to follow for algorithms to perfectly understand information searches is still long, although it now seems a little more downhill.
Should Websites Be Optimized for Google BERT?
As we have seen, Google BERT is going in the direction of getting a better understanding of content.
Conceptually, it is not possible to optimize a text from an SEO perspective for Google BERT: this technology, in fact, serves to understand the true meaning of the queries written by users and return more relevant results.
As Google's Danny Sullivan explains, BERT does not change the foundations of SEO.
If you are already working to create high quality content , then you are already optimizing it for Google. At the limit, you can think even more about people's real needs and better understand how you can respond to their needs. In this sense, we recommend the Google Trends guide by our SEO Federico Giannicchi.
It will be even more important to think about the type of search intent (informational, navigational and transactional) to produce the type of content that is truly relevant to the type of searches our audience performs.
The old saying that to rank you must produce excellent, high-quality content is still true, then! what do you think about it?
