What is the BERT algorithm? Basic information and necessary SEO measures

Search rankings are one of the concerns, as they can have a significant impact on sales for companies that use their websites for sales and promotions. Web staff who are focusing on SEO measures, what do you think about the BERT algorithm that Google updated in December 2019?

The BERT algorithm, which is one of the natural language processing technologies, has a different impression from previous algorithm updates. There are probably many people in charge of the web who have not yet considered specifically what to do with their homepage . In this article, we will explain what the BERT algorithm is and the necessary measures in an easy-to-understand manner.

What is the BERT algorithm (text-type search queries increased)

BERT is an abbreviation that stands for Bidirectional Encoder Representations from Transformers, and is a natural language processing algorithm that utilizes artificial intelligence announced by Google in October 2019 At the time of the announcement, it was targeted for English searches in the United States, but it was updated in Japan on December 10, 2019.

At this time, it was introduced into more than 70 languages, and the BERT algorithm began to be used worldwide. Google claims that it improved the accuracy of 10% of searches , demonstrating the power of the BERT algorithm.

What is natural language processing

Natural language processing is a branch of artificial intelligence and linguistics. Although it is processed by a computer while considering the ambiguity contained in human language, it was considered difficult to put into practical use because natural language differs depending on the environment of each person. Natural language processing first decomposes a sentence into words and analyzes the connections between each word.

Then, it analyzes the meaning of the sentence by linking highly related words . In this way, all sentences are analyzed, and the context is analyzed from the connections between sentences. Natural language processing, such as automatic translation and chatbots, is a technology required in modern society, and development is required.

What Makes BERT Better?

Conventional natural language processing uses a single natural language model for parsing and addresses a specific task. However, BERT can be pre-trained using a large amount of data available on the Internet . In addition, a new model can be created from learning data, called transfer learning, so it is possible to handle various tasks even with limited data *. *Conventionally, it was necessary to train each model to learn vocabulary, but BERT works efficiently because it can use data that has already been trained.

Sentence-type search queries are more relevant

BERT will be used in Google ‘s algorithm, and it will be possible to optimize users’ search intentions more than ever. Therefore, sentence-type queries such as natural questions will show the results of introducing BERT. *With the spread of voice search, it is believed that the convenience of user searches and the accuracy of results will continue to improve.

What can be done with BERT

How will the introduction of the BERT algorithm make it different from conventional searches? Here are some examples of what you can do.

Better understanding of context and improved accuracy

*The BERT algorithm can understand context from sentences containing words, not just words. *Natural language processing technology enables Google ‘s search engine to understand the language that humans speak on a daily basis . This means that search engines can understand what you want to know from context and display relevant search results .

Until now , search engines have been able to understand keywords, but not the context. Therefore, even if the content was not really what the user was looking for, it was associated with the search results due to keyword associations. The BERT algorithm allows Google to better respond to user search intent by understanding context .

Before and after introduction of BERT algorithm

Let me introduce a specific example that can compare before and after the introduction of BERT.

According to the article “Understanding searches better than ever before” published by Google on October 25, 2019 , even if prepositions such as “for” and “to” have important meanings, the BERT algorithm extracts the context. He mentions that users search naturally because they can understand it.

An example was given for a search for ‘2019 brazil traveler to usa need a visa’. In Japanese, it means “Travelers to Brazil 2019 need a visa”, where the word “to” plays a particularly important role in user search intent. This is an example of providing accurate search results to users by increasing the accuracy of reading context .


“to” indicates the important part of “Brazil to the United States”, but before the introduction of the BERT algorithm, the search results were interpreted as Americans traveling to Brazil, as shown on the left side (Before) of the above figure . I will return it. However, with the introduction of the BERT algorithm, it is possible to display content that states that Brazilians need a visa to travel to the United States, as shown on the right side of the figure above (AFTER).

As you can see from this example, the *BERT algorithm can flexibly respond to the user’s natural questions rather than keyword combinations. *It means that users can more accurately respond to their search intentions by searching with natural input such as conversation, instead of thinking about what the user wants to find in their head and searching by entering keywords themselves. Voice search, which has been increasing in recent years, is exactly such a search term. Looking at Google Analytics, many people may have noticed that many of the search queries feel like voice searches. The act of “searching” is becoming more familiar.

Does BERT need countermeasures?

What kind of measures should be taken by web content operators in the search environment after the introduction of BERT ? Explain what you need to do to get higher for users’ searches.

Create easy-to-understand sentences

While BERT makes it easier to understand context, you still have to create content that Google can understand . * Try to keep the title , headline, and body text as simple as possible, and be careful not to make any typos. *Once you have finished writing the text, of course, you should read it again, and if it is possible to double-check it, have someone other than the author check it.

Also, I sometimes see cases where a sentence is broken in the middle of the sentence and a line break is made. This is a common style of writing when writing on a PC, but when viewed on a mobile phone, line breaks occur in strange places and readability is low. It’s inconvenient for readers as well as Google , so be careful.

Optimize for user search intent

If your content doesn’t contain enough information for your users , it’s impossible to rank highly on Google . Make sure your content is better than your competing content .

Keep in mind that content that ranks high in search results is currently highly rated by Google . When creating content, you need to think specifically about what your users want to know and what motivates them to search Understand what your users are looking for and prepare relevant content to increase relevance to search queries .

Create content with voice search in mind

The background to the introduction of BERT may also be the activation of voice search, which continues to spread. *BERT is best suited for conversational queries. *In order to make the created content compatible with voice search, one possible countermeasure would be to prepare problem-solving content that shows answers to questions.

The time has come to demand even higher quality content

With the introduction of the BERT algorithm, users are now getting better search results than ever before . With support for voice search and more, Google ‘s algorithms are making our language even more comprehensible. In search engines with such specifications , it is important to improve the quality of content in order for users to find your content .

In a nutshell, it is all about creating content that is optimized for the user’s search intent with easy-to-understand sentences that can respond to sentence-type queries . Algorithm updates are a matter of concern for content operators about the impact on search rankings, but don’t be afraid to create content that users need.

Leave a Comment