Algorithm Rolled Out: October 22, 2019

Algorithm Overview: Overall Gist

Google’s upgrade is designed to aid in processing natural language by utilising an algorithm known as Bidirectional Encoder Representations from Transformers, or BERT. The BERT AI update is intended to advance the science of language comprehension by applying machine learning to a large corpus of text – in this example, a Google search. They believe it is the most significant change to their search system in five years. The most significant modification to the search algorithm since Google’s renowned RankBrain upgrade in 2015.

The BERT search query algorithm evaluates words typed into Google concerning all other words in the phrase, rather than one at a time. The AI is then used to rank and highlight snippet results to locate results for users more precisely. Google believes that the BERT model will affect one out of every ten searches in the United States. Its capacity to interpret conversational English means that it can now grasp the context that prepositions such as “for” and “to” offer to a word.

More specifically, the BERT language processing model can contextualise each word in a phrase about every other word in the phrase at the same time. This distinguishes it from natural language processing (NLP) models that contextualise words in left-to-right or right-to-left order.

It accomplishes this in part by “masking” some of the words in the input text and then having the BERT model predict the masked word bidirectionally. This allows it to deduce the meaning of each word based on the context of the language and prevent misunderstanding with synonymous words.

What is The BERT Algorithm Update?

BERT is Google’s newest neural network-based approach for pre-training natural language processing (NLP). BERT (Bidirectional Encoder Representations from Transformers) is an abbreviation that stands for Bidirectional Encoder Representations from Transformers. Let’s dissect the acronym.

Bidirectional – Previously, Google and many other language processors just looked at requests in one way, either right to left or left to right (usually depending on the language being searched). However, as we all know, most of what we do in language necessitates analysing everything stated in relation to everything else in a phrase. The “B” in Bert represents the Natural Language Processor, which assists BERT in understanding how different portions of the query connect.


These are words that serve as adhesive words in a phrase that Google’s search algorithm has mostly overlooked. Previously, when someone searched for “parking on a hill with no kerb,” Google looked at the key phrases “parking,” “hill,” and “kerb.” Google already rejected the crucial transformer: “no.” As you can see in the screenshot below, the old Google search would provide SERP results for parking on a slope with a kerb while disregarding transformers. When BERT added to the existing Google search algorithm, the search engine returns more relevant results to the user.

BERT learnt how to detect individual words and what those words signify based on how they combined and the context in which the words are used. BERT is a significant advancement in how machine learning handles natural language.

Essentially, BERT is attempting to comprehend more than just our words – it is attempting to grasp the purpose of those words, making it critical to the future of search.

Because BERT relies on natural language and text context, paying attention to the surrounding context of keywords might be one method to optimise for this new model. Creating on-page content that focuses on subjects rather than keywords will maintain readers’ attention to the site’s content. Furthermore, site owners should concentrate on creating high-quality material that adheres to Google’s best practices criteria and has a high degree of EAT (Expertise, Authoritativeness, Trustworthiness).

Where has BERT been rolled out?

While BERT initially exclusively utilised organic search results from, as of December 2019, it has expanded to more than 70 languages globally. BERT has previously utilised for Featured Snippets, displayed at position 0 in organic search results with text, table, or list, in all 25 languages in which Google shows Featured Snippets.

BERT is now available in the following languages for the calculation of organic search results: Afrikaans, Amharic, Arabic, Armenian, Azeri, Basque, Belarusian, Bulgarian, Catalan, Chinese (Simplified & Taiwan), Croatian, Czech, Danish, Dutch, English, Estonian, Farsi, Finnish, French, Galician, Georgian, German, Greek, Gujarati, Hebrew, Hindi, Hungarian, Icelandic, Indonesian, Italian, Japanese, Javanese etc.

Algorithm Winners & Losers





Algorithm Solution: Ways to implement or take to cope with Google algorithm guidelines

The additions of BERT and RoBERTa pave the stage for even more sophisticated achievements in the next years. SEO practitioners will now need better understand a whole new set of words and processes to equip their websites for the future of search.

  • Synthetic Data
  • GLUE (General Language Understanding Evaluation)
  • SQuAD (Stanford Question Answering Dataset)
  • SWAG (Situations With Adversarial Generations)
  • Greedy Decoding
  • Label Smoothing
  • Loss Computation
  • Data Loading
  • Iterators
  • BPE/ Word-piece
  • Shared Embeddings
  • Attention Visualization

There is no straightforward way to respond to BERT. There are no quick fixes that you can apply to boost your website’s ranking or recover losses. Instead, keep in mind that you should produce your content and construct your websites for people, not algorithms: for your future users and customers who will be browsing and engaging with your website.

Real-Time Implementation Example

Context and BERT in Action:

The phrase was “how to catch a cow fishing?

The term “cow” in the context of fishing in New England refers to a huge striped bass.

Striped bass is a popular saltwater game fish that are caught by millions of anglers throughout the Atlantic coast.

So, when doing research for a PubCon Vegas talk earlier this month. I searched in “how to catch a cow fishing,” and Google returned items pertaining to animals, specifically cows.

Despite the fact that I had used the word “fishing” to add context. Google ignored it and returned results about cows. On October 1, 2019, that happened.

On today’s date, October 25, 2019, the same inquiry returns a plethora of striped bass and fishing-related results.

The BERT algorithm appeared to recognise the importance of the context of the term “fishing”. And redirected the search results to pages on fishing.

BERT has an impact on which search queries.

The influence of BERT on long-tail search queries is significant. BERT improves the interpretation of context for longer queries submitted. (or spoken for Voice Search) as a question or a series of words into the search field.

Google gave a few examples of search questions that BERT aids in better understanding. And for which the search engine now returns more relevant results on their blog.

According to Google, the importance of the word “to” and its link to other terms was previously underestimated in this example for an organic search result. The word “to,” on the other hand, is crucial to the meaning of the statement. 

We’re dealing with a Brazilian who wants to visit the United States, not the other way around. Google can now correctly interpret this distinction thanks to the new BERT model, and give results that match the genuine search meaning.

BERT vs. RankBrain: What’s the Difference?

BERT’s capabilities are similar to RankBrain, Google’s initial artificial intelligence system for analysing searches. However, there are two different methods that could utilised to help with search results.

“The first thing to realise about RankBrain is that it works in tandem with traditional organic search ranking algorithms, and it’s used to tweak the results calculated by existing algorithms,” said Eric Enge, general manager of Perficient Digital.


Final End Result

Google has always considered it a priority to serve its consumers with better, more accurate search results. As a result, its algorithm has become increasingly sophisticated in understanding language and searcher intent — first with RankBrain and now with the BERT AI upgrade. The Google BERT upgrade focuses on natural language processing (NLP) to assist the search engine in better understanding previously unseen terms and queries, and it’s getting closer to that aim. It is critical to remember that your material always aimed at the same general audience: people. RankBrain and BERT algorithm upgrades simplify for Google to comprehend the intent of a query based on linguistic characteristics. This trend is expected to continue in future algorithm updates.

0 CommentsClose Comments

Leave a comment