SEO Cape Town Guide: All you need to know about The Google BERT update
In this SEO Cape Town Guide, we will be looking into the BERT update, how it affects our search, and what it means for search overall. This twelve-minute read will guide you through the ways in which BERT has impacted search and what the future of SEO will look like as a result of this change.
Towards the end of 2019, Google made an announcement about a huge update with regards to how algorithms work that would impact on 10% of all Google searches made. This update is the introduction of a major language processing tool called BERT, which uses technology to accurately assess the ways that words in searches relate to each other, allowing Google to develop a greater understanding of the intention behind the search.
With its neural network-based technique for natural language pre-training, BERT’s concept is to train computers to understand the concept of language in a similar way to that which humans do.
This could be seen as one of the most powerful developments in algorithm technology over more recent years.
SEO Cape Town Guide Outlines BERT
To properly understand the BERT, update you must first understand what it means. BERT stands for Bidirectional Encoder Representations from Transformers.
The technology behind BERT allows anybody using search engine technology to train their own, bespoke question answering tool. This concept came about as a result of Google’s in-depth research into transformers- or models that process all of the words in a search in relation to the other words included within that particular sentence- not just singular words, or words written in a particular order as a complete sentence. Google created a technology that would consider the bigger picture- the full context of a word, by taking into consideration the words that come before and after that keywords- therefore beginning to understand the searcher’s intent behind their search queries.
Google knew that this development would have to go beyond the advancements in software in order to bring the update into ultimate fruition- they also needed to develop hardware which could cope with these advancements and would deliver search results quickly getting you the most relevant information in the fastest way possible.
The SEO Cape Town Guide to Working With Real-Life Queries
By applying this BERT framework technology to ranking and featured items found within search, Google’s aim is to be able to find a better way of helping searchers to source more relevant information, relating directly to their searches. This has now directly improved up to one-in-ten searches in American English and is set to be rolled out to be used with other languages and nuances over time.
Search will now begin to understand the context of words that are used within a search query. Something that is particularly useful when searching for conversational, longer queries where words such as ‘for’, ‘to’ and ‘from’, which can influence and alter the meaning behind a term quite significantly. This allows for a searcher to input words that feel more natural to them, while still receiving the expected similar result from their search queries.
To make the best of this technology and to launch the improvements to search, Google undertook vigorous testing to ensure that these changes would prove to be more helpful to users.
SEO Cape Town Guide: Examples of BERT in Action
In it’s testing phase, Google used the search term “2019 brazil traveler to usa need a visa’ as part of the process. In this context, the use of the word ‘to’ and its proximity to the other words within the search term are important in order to begin to understand the meaning behind the word.
As humans, we are able to compute that this term is about a Brazilian traveller going to the United States of America- not the other way around.
Groundbreaking technology means that this formerly misunderstood search term– returning results about United States Citizens’ travelling to Brazil, now uses BERT to grasp the nuances behind the search in the same way that humans are able to compute. BERT is now able to understand that the commonly used word ‘to’ plays an integral part in returning accurate results to a searcher’s query.
(source: Google Blog)
Another query used within the testing phase was “do estheticians stand a lot at work’’. Former systems would take a word matching approach to understand search terms, filling in the gaps of what the searcher’s intent could be rather than what it actually was. They would match the term ‘stand-alone’ in the query, rather than the term ‘stand’ on its own, which once again altered the meaning behind the search term. Now, BERT framework models would understand that ‘stand’ would mean the concept of physically standing, which relates to the job of an esthetician, and now produces a much more useful response to the term.
Google is now beginning to apply to make the search experience better for various written languages across the globe. One of the most powerful characteristics of this model is that they can take the things they have learnt from developing their understanding of spoken and written English. The SEO Cape Town Guide to Understanding Linguistics.
While developing the BERT model, Google had to delve into the world of linguistics in order to better understand the way words fit together and why. The natural understanding of language can be linked back to the Turing Test Paper of over 60 years ago.
Lexical ambiguity brings with it many unsolved problems relating to the ambiguous nature in which the English language holds- one of the most complicated issues being that nearly half of the words in the English language can mean something else, which is, naturally, a challenge for SEO, in producing the ever-growing content for search engines who try to interpret the meaning of search terms and content to meet the needs of users in their queries.
It begins at the sentence rather than a word-for-word level, combining words with more than one meaning to make a sentence- creating phrases that are complicated to understand with some of the specific rules of linguistics playing their part in the complicated web of language and search.
Examples are- Polysemy and homonymy: words that have multiple meanings, such as ‘rose’ (colour) ‘rose’ (flower) and ‘rose’ (getting up).
However, the complication can run even deeper than this. Polysemous words have two or more meanings, but their origins are subtly rooted in the same place. For example, ‘get’ (verb) could mean ‘to retrieve’ or ‘to comprehend’, similarly with the word ‘bass’- could be either an instrument, a tone, or even a fish.
SEO Cape Town Guide to Lexical Ambiguity in Humans and Machines
Through accent and pronunciation, humans are often pre-conditioned in their language acquisition to be immune to these challenges. As the conversation develops and continues, we understand who ‘they’ or ‘it’ is being referred to when reading sentences, or even participating in a spoken conversation as we as humans are able to keep a track of the subject being spoken of.
However, machines do not necessarily have this pre-conditioned intelligence when it comes to understanding contextual clues in conversation. They do not understand that somebody is fishing at the river bank, while somebody else is depositing money into the bank. Unlike human beings, machines can lose track of the topic of conversation (search terms) which is a challenge within search.
Human beings will realise that when a sentence contains related words to the above such as ‘water’ ‘nature’ ‘outside’ and ‘fishing rod’ that the bank in which is being referred to is not the financial establishment, but indeed, a riverbank. We are acutely attuned to the context in which a word is written or spoken in a way that makes cognitive sense to us, making the handling of nuance and ambiguity something that comes naturally to us.
With machines, conversational search adds complication to the mix, in particular, when searching for longer, more complicated sentences and phrases together.
The SEO Cape Town Guide to What This Means for the Future of SEO
Many SEO experts may not have even noticed the Google BERT update, and for good reason. The update was designed to tackle longer and more complicated search terms with a conversational tone, which would not have impacted so much on generic, more concise search terms. BERT’s impact on developing a deeper understanding of conversational search terms would not have played as big a part within SEO. When looking into these rankings, experts may not have kept track of these longer search queries- instead of focusing on those which send a greater value of traffic towards your website, which tend to stay on the shorter side.
The SEO Cape Town Guide to Making The Best of The BERT change
The BERT update represents a refining of the algorithm rather than a wholesale change, which means it will be difficult at first to see the profound impact it has. However, it does reinforce what SEO Cape Town has preached about SEO for years — being conversational is key. Knowing the right target keywords or identifying the right trending keywords is only the first step in strong SEO.
A lot of people who work in the field stop there, frontloading top keywords without asking if they are the best keywords and not prioritizing the structure of their use, whether in the metadata or a headline. But the best route to seeing your content surface in results is to keep the human factor in mind. It’s not just that you want the content to surface, you want the potential audience to feel enticed to click — to know that your content is the answer to the question they are asking.
With that wisdom in mind, conversational becomes key. Within SEO, keeping the best keywords front and centre while remaining conversational has always been the fine line one had to walk. It is an imperfect process, but it is data-supported and a skill one can build.
The BERT update leans into this by factoring in the complexities of language to allow for greater flexibility in being conversational with SEO. This is important because of all the way the SEO algorithm collects information when making a ranking decision — metadata, headlines, subheads, related links, inline links and more are all factored into how a site ranks. The more Google can determine SEO, the higher the wealth of knowledge on a topic, the better SEO sites will rank.
The more conversational you can be, the more likely a person will want to click on your site. BERT marries the concepts together a bit more smoothly, if in a very subtle manner. The more conversational one is with their language, the more likely it is for linguistic nuance to cloud the searcher’s intent and muddy the waters when it comes to the search results. This is especially true in English, where the language lends itself to an entire host of colloquialisms and multiple meanings to words. The update is designed to really drill down to determine searcher intent in even the more conversational phrasing of key terms.
In a way, it’s taking the context of being conversational with SEO on the results side and applying it to the searcher intent side of the equation. The algorithm change not only rewards more conversational keywords than ever before, but it also accounts for more conversational approaches to searches by users than ever before. It’s a small distinction, but, in time, a critical one.
The more the algorithm evolves, the more the SEO process mimics genuine conversation and communication. They ask a question, we provide an answer. The way that looks will increasingly look like dialogue that could have existed in the real world, which can streamline thinking when it comes to SEO but it also allows for creativity and innovation.
So, while BERT is a subtle innovation within search theory, it is almost certainly an indicator of a more lengthy, profound evolution in SEO.
Trending SEO Cape Town work lives on the bleeding edge
To see the impact of this evolution in real-time needs to look no further than the evolution of trending work in the news industry.
During big breaking news stories, sporting events and more, news agencies are learning how to wrap their heads around real-time search trends in order to determine searcher intent. In other words, they are identifying massive spikes in search interest around particular keywords and then attempting to determine exactly questions is being asked around those keywords and answering it before people move on to another question.
The windows for these efforts tend to be quite small, sometimes amounting to minutes rather than hours, but hitting the mark when it comes to matching keywords to searcher intention can mean tens of thousands of page views for a site.
The BERT update leans into this idea because trending work by nature has to be fun, conversational and it has to be a clear answer to the question.
For example, If someone is watching a football game and a player gets hurt on a live broadcast, audiences are going to immediately begin searching for news around the injury. Do they want an update? Did they miss the game but want to see the play that caused the injury? How are people reacting to the injury? What does it mean for future games? Is just one of these questions being asked or are all of these questions being asked.
Trending writers have mere minutes to not only determine the question but also fashion the answer in a clear, conversational way that makes it clear that they recognise the question and the answer is within the story that they produce.
In fact, strong trending SEO also centres around the concept of joining a conversation on the whole. If a popular awards program is on television, for example, and a celebrity is wearing an insanely bizarre outfit, rest assured there will be a host of people online looking to see if other people thought it was a bizarre as they did.
This second-screen opportunity represents a core idea behind conversational SEO and the opportunity created by the BERT updates — connecting humans to humans. Think of times you may have done this yourself. What did you see? What questions did you ask? How was that reflected in your search? Where did you find your answer and why did you choose it?
At its heart, that is what the algorithm is, what it does and what it’s for — human connection. It is the greatest facilitator in helping people find answers together. Its power is its simplicity. It is used by almost everyone, every day and its strength is that it’s so effective that you don’t think about there being an intermediary in your search for answers at all. The BERT update cuts through linguistic nuance to improve that facilitation, representing a great leap forward and a hint of what’s to come.
The more that people can utilise SEO in a way that sounds like humans talking to each other while respecting the data-intensive power of keyword rankings and how it can be used to elevate a site’s position, the more long-term success that can be achieved in the wake of the BERT update.