Google algorithm changes normally strike a sense of dread or at the very least a sense of trepidation into the heart of all content producers. Previous updates such as hummingbird, penguin and panda have seen drastic changes in the way the web is portrayed, largely benefiting the end-user. However, the recent update to the algorithm duly dubbed BERT, Bidirectional Encoder Representations from Transformers (you can’t make this stuff up), is the first that genuinely signals a seismic shift in how the end-users are already using Google rather than how they could.

BERT is a machine learning algorithm that aims to better understand your search queries and match them with the content on a page. This algorithm is optimising the search results of long-form queries using neighboring words to contextualize others. Below you can see a classic long-form query someone might be searching for regarding obtaining a visa. In the first search before the BERT algorithm, you can see how Google has provided an up to date news article from a reputable source on US citizens traveling to Brazil without the need for a VISA. Whilst interesting for the end-user it hasn’t answered their question. In the second query, you can see how Google has identified the intent of the user to purchase a visa through contextualising the query by identifying ‘need’ as an important element.

Bert Example

Whilst this will have little to no effect on the end-user it does change how content producers are writing pieces. BERT signals the move towards prioritising human-centric pieces, whilst before authority was deemed the most important aspect when prioritising websites now Google is starting to look at how best to answer your queries. As such priority number one for writers should be to make sure they are making original clear content. Whilst this is still in its infancy this move is coming and marketeers need to be aware.

Natural Language Processing In BERT

So what does this signal in the changing habits of the end-user? Most notably that people are starting to treat machines like they would humans. Our expectations are that google should easily be able to understand what I am trying to ask and provide me with the outcome that I am looking for. It wasn’t to long ago that if you were looking for a specific article or web page that you would have to type in a series of disparate keywords in the hope that Google would match them with your sort after destination, often having to perform a couple of searches to refine this down. These days our searches are becoming ever closer to natural language and it is now up to search engines like google to catch up rather than them dictate how you use them. It is now much more about the user’s expectations rather than being taught how to use a tool.

Whilst this has been spurred by the increase in search query length, with searches of 6+ words now making up 12% of all internet searches (https://www.slideshare.net/randfish/keepng-up-wth-seo-n-2017-beyond/11-Keyword_Length_of_Search_Queries1Word) over the next 2 years we are going to see this increase exponentially. Driving this is the ever-increasing use of AI assistants, using voice search as their main way of answering queries. As of 2017, OC&C Strategy Consultants reported that 17% of households in the united states owned a smart speaker with this number predicted to rise to 55% by 2022. This just shows how important natural language processing is going to be in the future.

So What does the Future Hold?

With technology aligning closer to human nature every day our expectations are only going to get higher. 5 years ago it would be absurd to ask an AI assistant ‘Book me a table at [Insert restaurant] for 19:00 Friday’ but now we are annoyed when it doesn’t get this right first time. This way of thinking is also dripping into every aspect of our lives, it seems that a barrier has been crossed where we expect a certain level of intelligence from automation, whether this be the ads we are viewing, the searches we are conducting or posts in our social media feeds.

Whilst BERT marks the start towards natural language processing in search, in reality, this is a small step and something that I expect signals the start of a massive movement. At EngageIQ we are harnessing natural language processing to understand end user’s sentiment in our research. By transcribing our conversations, we can then run BERT to try to add sentiment and ultimately understand the end-users goals. As expectation changes how we market to end-users will have to change overall mediums, outbound marketing efforts have to be personalised, with the end-user wanting salespeople to understand their own sentiment before speaking to them.