Skip to main content
Coherence logo Coherence logo

Articles /

Big changes to Google’s search algorithm from BERT

31

Oct 2019

Julian Tedstone

Managing Director

At Coherence, we are long standing advocates of Sarah Richard’s work on Content Design. This clear-eyed look at how we communicate with our users brings some home truths. One nugget is to not confuse the literacy of the user with their intelligence. UK Government web sites have aimed web content at the literacy level of a 9 year old (for citizen content, 14 for corporate) and that an average 12 year old has a vocabulary of 50,000 words, for example.

This does not mean that the average adult has the knowledge, life experience or judgment of a 12 year old. The two groups might share vocabulary and syntax, while still using them in very different ways. 

As people who do things on the web, we should do our best to make our content as accessible as we can. This means using language that is the most inclusive. It also means working hard to understand the intent of our users' language. Why wouldn’t we? 

Stuff keywords

So, why do we stuff keywords into our web content? We’ve known for a while that Google will prioritise ‘natural’ language in web copy. But there is still the temptation to jam-pack that content (and its meta content) with lots of rich keywords. It’s as if we need to create a nectar pot to attract the bees. The more nectar, the more bees. Except this does not always flow through to conversions (you know - people doing something that shows your web content is useful to them). This is because humans talk like humans and not like search optimisation techniques.

Say hello to BERT

Give a cheery hello, then, to BERT, something new that Google has been cooking:

“Last year, we introduced and open-sourced a neural network-based technique for natural language processing (NLP) pre-training called Bidirectional Encoder Representations from Transformers, or as we call it--BERT, for short. This technology enables anyone to train their own state-of-the-art question answering system.”

Understanding searches better than ever before, Pandu Nayak, Google Fellow and Vice President, Search.

None the wiser? Well, to this bear with a very little brain, I think it means something like this:

Searches typically look for keywords - Big Important Words that carry a Lot of Meaning. At school we called them content words (important nouns, verbs and sometimes adverbs and adjectives). Nouns that denote special things (places, people, Things In The Real World That Exist) get noticed in a search term. So a search term like ‘Where are the best restaurants in London” will have an obvious focus on Location and Thing. Boring words like in, on, next to (prepositions etc) or connectors (and, or, but, with) tend to get ignored.

So the thinking goes; the more we can stuff our content with Big Important Words, the better visibility we shall get for our website. Except users don’t talk like paragraphs stuffed with keywords. And our searches are more sophisticated and contextualised than this - even when they are using the syntax and vocabulary of a 9-14 year old.

Big little words

And it turns out the little words (preps and convectors) are key to understanding intent. Consider this great example from Google. They can explain better than I can:

“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 U.S., and not the other way around. Previously, our algorithms wouldn't understand the importance of this connection, and we returned results about U.S. 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.”

Understanding searches better than ever before, Pandu Nayak, Google Fellow and Vice President, Search.

So a little word like to can now deliver a big change in meaning - in effect telling us the intent of the user's search.

TLDR?

In summary, write like a normal person.

Julian Tedstone

By Julian Tedstone
Managing Director

Share this