Google’s latest update has SEO experts in a state of confusion. Google BERT is an improved version of the Google Rankbrain algorithm . Google considers this update as “one of the biggest advancements in the history of search engines.” But from my point of view, the impact of the algorithm would actually be minimal.
What to think then? The BERT algorithm is an open source neural network that enables automatic processing of natural language. The acronym stands for “Bidirectional Encoder Representations from Transformers” and was introduced by Google last year.
BERT is a technology that allows word integration and disambiguation through word embedding. This is its main quality, but unfortunately it is also its biggest flaw, since this technique requires a large computing capacity and its use on a large scale is therefore quite expensive.
Word2vec and other models allow word vectors to be precomputed for each sentence and stored in a database, so that they can be extracted later. This is not the case with BERT, which requires systematic recomputation of the vectors.
What is the difference between word embedding (word2vec) and BERT?
Here’s an example:
The lawyer meets with his client several times before the trial.
Sentence 2: To choose a good avocado : pay attention to the color and appearance.
If I want the algorithm to understand the meaning of the word “LAWYER” which designates both a profession and a fruit, with word embedding, it is enough to extract THE pre-computed vector from the database. With this approach, there is only one vector for the same word.
In contrast, BERT is a Telemarketing Leads technology that creates “contextualized” vectors, so I will have to analyze the two sentences in the BERT network. BERT will generate two very different vectors for the word “lawyer” that appears in two very different contexts. We are then talking about lexical disambiguation.
Thanks to its disambiguation ability, BERT is at the forefront of progress in many natural language understanding tasks. However, this ability also makes it computationally demanding and therefore difficult to exploit. This is why this type of algorithm is powerful in question analysis or content rewriting. However, it has its limitations.
Its computational requirements make its application in search queries limited, at least for now. There is no doubt that there will be improvements in the near future and that the processing needs will eventually be reduced.
How does BERT help Internet users?
Google’s BERT update allows Internet users to obtain better results from long conversational queries (Google is always oriented towards voice search optimization ). Today, it is no longer necessary to use an unstructured sequence of keywords to be understood by Google.
If the Google BERT update directly affects Internet users, it is more complicated when it comes to content specialists. Google Hong Kong Lead has improved its understanding of context, but that doesn’t mean you should start writing thousands of pages aimed at the long tail.
This update rather encourages us to continue writing well-organised, rich and complete content.
The impact of this update is not unanimous. Google claims it is the “biggest leap forward in the past five years” with 10% of search results affected by the change.
