BERT, introduced by Google on October 25, 2019, is a natural language processing technology. Initially limited to English-speaking regions, it was later implemented in Japan on December 10, 2019.

The BERT update caused quite a stir and was hailed as one of the most significant updates, garnering considerable media attention.

It was estimated that the BERT update affected about 10% of search results. However, due to its ability to return search results with high accuracy even for conversational and lengthy queries, some have noted that SEO optimization has become easier as a result.

Tokyo SEO maker will explain what BERT is, how it has changed with the update, and what future SEO strategies should be entailed.

What is BERT?

BERT, short for Bidirectional Encoder Representations from Transformers, refers to representations that can be encoded bidirectionally using a deep learning model called Transformer.

*Encoding refers to converting data into a different format.

BERT is a type of natural language processing technology (NLP), which refers to the technology that enables computers to understand human language.

In other words, Google’s implementation of BERT in its search engine means that it can now better understand human language and represent it in search results.

Understanding Text with BERT

One of BERT’s significant features within natural language processing technology is its ability to understand context. Previously, Google’s algorithms would simply return search results based on the entered keywords. However, since the adoption of BERT, Search now returns results based on understanding the context of the text and providing more relevant results that satisfy the user.

For a clear example, Google’s blog post, ‘Understanding searches better than ever before ,’ explains as below.

“Here’s a search for “2019 brazil traveler to usa need a visa.” The word “to” and its relationship to the other words in the query are particularly important to understanding the meaning. It’s about a Brazilian traveling to the US, and not the other way around. Previously, our algorithms wouldn’t understand the importance of this connection, and we returned results about US citizens traveling to Brazil. With BERT, Search is able to grasp this nuance and know that the very common word “to” actually matters a lot here, and we can provide a much more relevant result for this query.

The gist is that, when searching for ‘2019 brazil traveler to usa need a visa,’ the search results should ideally show ‘a Brazilian traveling to the US’ However, before the BERT update, results showed ‘US citizens traveling to Brazil.’

The original search query was ‘Do Brazilian travelers need a visa to go to the US in 2019?’ Previously, the search results returned were for ‘American citizens traveling to Brazil.’ However, after the update, the results now correctly reflect the original query.

This example demonstrates how BERT correctly understands the context of the search query, leading to accurate search results.

BERT is an example of natural language processing technology that excels particularly in understanding context, including grammar.

The Reason Behind BERT’s Implementation

The reason for implementing BERT is to correctly understand search queries and deliver search results that satisfy users. In other words, it’s a result of prioritizing user-first principles.

However, another reason for introducing BERT is speculated to be its adaptation for voice search.

While the adoption rate of voice search is not very high in Japan yet, it is significantly prevalent worldwide. According to ComScore, it’s predicted that by 2020, 50% of searches will be voice searches.

As a side note, the data claiming that 50% of searches would be voice searches by 2020, as mentioned by ComScore, has not been publicly disclosed. It’s speculated that this data may have been somewhat erroneously propagated, as back in 2014, it was known that 10% of searches on Baidu were voice searches. There was a prediction that if this trend continued, 50% of searches would be voice or image searches within five years.

Reference:

-STOP Using Comscore’s 2020 Voice Search Statistic (It Never Existed!)

-Inside Baidu’s Plan To Beat Google By Taking Search Out Of The Text Era

-The impact of voice assistants on consumer behavior: PwC

The proliferation of voice search is a correct speculation, and accurately understanding human language is essential for returning accurate results.

As the use of voice search systems like Alexa, Android’s ‘OK Google,’ and Apple’s ‘Hey Siri’ advances, there is a growing trend towards spoken language searches. This is another reason why search engines that can understand context correctly have become necessary, which may have contributed to the implementation of BERT.

Changes in Search Results due to BERT Implementation

Apart from the Brazil example mentioned earlier, Google’s official blog ‘Understanding searches better than ever before‘ provides four other instances

Search Query: do estheticians stand a lot at work

“Previously, our systems were taking an approach of matching keywords, matching the term ‘stand-alone’ in the result with the word ‘stand’ in the query. But that isn’t the right use of the word ‘stand’ in context. Our BERT models, on the other hand, understand that ‘stand’ is related to the concept of the physical demands of a job, and displays a more useful response.

According to Google, before BERT, the search results for the query ‘do estheticians stand a lot at work?’ were misunderstood, associating ‘stand’ with ‘stand-alone’ and returning results related to isolation. However, BERT has improved this understanding.

Search Query: Can you get medicine for someone pharmacy. 

“With the BERT model, we can better understand that ‘for someone’ is an important part of this query, whereas previously we missed the meaning, with general results about filling prescriptions.

Previously, we displayed general prescription-writing methods in search results, but now, we can correctly recognize the crucial part of the phrase ‘for someone pharmacy.

Search Query: parking on a hill with no curb. 

“ In the past, a query like this would confuse our systems – we placed too much importance on the word ‘curb’ and ignored the word ‘no’, not understanding how critical that word was to appropriately responding to this query. So we’d return results for parking on a hill with a curb!

While we understood ‘hill’ before, we failed to understand ‘no curb’ and only recognized ‘curb’, resulting in simply returning parking methods for a hill. After the improvement, the results are more aware of whether there is a curb or not.

Search Query: math practice books for adults

“ While the previous results page included a book in the ‘Young Adult’ category, BERT can better understand that ‘adult’ is being matched out of context, and pick out a more helpful result.

It’s a query for math books for adults, but previously, we emphasized ‘adults’ and displayed math books for middle school students. However, BERT has improved this.

No Need for SEO Strategy for BERT Update

With every Google core algorithm update, there is often information circulating about how to address the update, but as the improvement with BERT update lies in its ability to understand text correctly, there are no improvements to be made through SEO strategies.

Summary

SEO experts suggest that if rankings have dropped since the BERT update, it’s likely because the search intent wasn’t being correctly fulfilled before. Therefore, it’s more reasonable to interpret that the update made correct judgments. Hence, rather than implementing specific measures, consistently creating user-friendly content with a user-first approach remains the best strategy to address updates.

Author Profile

Mr. Takeshi Amano, CEO of Admano Co., Ltd.

Mr. Takeshi Amano is a graduate of the Faculty of Law at Nihon University. With 12 years of experience working in the advertising agency industry, he discovered SEO and began his research during the early days of SEO. He self-taught and conducted experiments and verifications on over 100 websites. Using this expertise, he founded Admano Co., Ltd., which is currently in its 11th year of operation. Mr. Amano handles sales, SEO consulting, web analytics (holding the Google Analytics Individual Qualification certification), coding, and website development. The company has successfully managed SEO strategies for over 2000 websites to date.

Return to the top of Japan SEO