In the wise words of Google, ‘search is not a solved problem’. Try as they might, search engines still don’t have a full grasp on human language yet – but they are getting closer.
Google’s latest systematic change to search, to better understand search queries, is one of the most meaningful updates in years – but what does this mean for your content?
Natural Language Processing
Google is consistently striving to build models that will help search engines understand language in the way that humans naturally use it, which is known as Natural Language Processing (NLP). NLP can analyse the syntax of a page to gain a better understanding of the sentences and also the context of specific phrases. Rather than searching for exact keywords to find what we need, we can ask long, fragmented or just plain weird questions and Google can still yield the results that we’re after.
In fact, with 70% of Google Assistant voice queries being made in natural language, this type of search lends itself perfectly to voice recognition technology, which has become increasingly popular over the last few years.
As well as being able to analyse the syntax, Google can analyse content as a whole to determine whether or not it’s well written from a topical perspective, as well as if it’s grammatically correct.
What is Google’s BERT update?
Enter BERT – or Bidirectional Encoder Representations from Transformers – which Google has cited as ‘the biggest leap forward in the past five years, and one of the biggest leaps forward in the history of search’. So, what is it about BERT that’s so ground-breaking?
Quite simply, it allows Google to better understand words in the context and the meaning of search queries. The algorithm uses pattern recognition to understand how people actually communicate, so that it pulls more relevant results.
To give an example that Google provided to demonstrate this update, if someone were to search ‘can you get medicine for someone pharmacy’, as humans we can infer that you’re asking if you can pick up medicine on behalf of someone else. However, before BERT, Google would previously have assumed it meant for yourself.
Another example would be if you searched ‘2019 brazil traveller to USA need a visa’. We know that, contextually, the ‘to’ and ‘need’ would mean that the person conducting this search needs a visa to visit the United States, but previously the top-ranking results would be about US citizens going to Brazil. Now, thanks to BERT, the top results would be for the US Embassy.
Semantic search is nothing new, and this isn’t a sudden change, as Google regularly updates its algorithms to keep pace with how the search engine is being used. One notable language update, Hummingbird, which was launched in 2013, took steps to understand the wider context and intent behind searches. Rather than just typing in keywords, users could start asking questions using more natural language. Google also has an NLP API tool that uses machine learning technology to pull insights from unstructured data and allows developers to apply natural language understanding to their applications.
Google’s goal has always been to deliver the best and most relevant results to the user, and users want answers fast, so understanding the intent behind those searches is key. Why are they searching? What are they hoping to find? Do they want an exact answer to their query, or are they trying to find a specific website? Voice search is a particular driving force behind this need for change, as people start to move away from keywords and more towards conversational enquiries.
BERT and Full Coverage
As well as trying to offer the most relevant search results, Google is starting to leverage BERT and one of its AI models to make sure that the results are reliable, too. In a ‘fake news’ era, Google is using them as a way to better understand if stories that are shown in the Google News full coverage area are factually accurate and reliable.
In short, this means that Google can now use the connections between articles and its fact check database to match facts more effectively with stories, as it can understand if the fact check is related to the main topic of the news stories.
How to optimise for BERT
It’s estimated that around 10% of all search engine results will be impacted by the algorithm update, so how can you make sure your content is ahead of the curve?
The truth is, there’s not a particular way to react to the update, as it wasn’t created to penalise specific sites, rather return more accurate results for users, so Google has stated there’s not really a way to optimise for it.
However, as the search engine is getting better at understanding our language, it’s more important than ever to write for a human audience. No keyword stuffing, adding in synonyms or any other tricks, just well-written, regularly refreshed copy that answers the questions that people are searching for. Keywords are still important, of course, but the content around them must be natural and cohesive.
There are still technical factors involved in search ranking, such as page authority and including internal links, but ranking for search queries now looks a lot more like understanding what your audience is going to be searching for, and writing quality content that answers their questions.