Google, with its search algorithm, uses an artificial intelligence ‘machine’ system called “RankBrain”.
How does it work and what do we know about it?
What is RankBrain?
RankBrain is a machine learning system (based on deep learning) developed by Google to improve the relevance of search results. It makes it possible to interpret new requests which have not been looked for previously.
The first official statement from Google RankBrain was made by Greg Corrado in October 2015. Greg said that soon after deployment, RankBrain became the third most important search ranking “factor”.
Given the impact, it is important to understand exactly how the algorithm works, and what it will bring to Internet users, as well as determining its impact on SEO.
How does RankBrain work?
As mentioned above, the main purpose of RankBrain is to provide the most relevant results by interpreting the meaning of whole sentences, rather than focusing on the keywords. One of the facets of this algorithm is the assimilation of content (this article will only deal with this aspect).
Rankbrain efficiently manages long-tail requests and their similarities to each other (How are they connected? To which particular subjects?).
Thanks to this, Google can understand a sentence that has never been treated before, by correlating it (via similarity) to already known sentences / concepts.
As a machine learning system, RankBrain is constantly learning, paying close attention to parameters like pogosticking, bounce rate, or time spent on a page. If a user considers the results displayed unrelated to the search, the algorithm will, the next time, display other results for this query.
What are the implications for Google users?
- You can find information on a concept or an event without using a particular or specific word (“Which animal is at the top of the food chain?”)
- You will also get more relevant results for ambiguous searches that have several meanings (for example, “Jaguar” like the car brand or “jaguar” like the animal)
- If you perform a search that Google has never encountered before, it will be correctly interpreted and adapted to the best known vector
Google doesn’t disclose the exact algorithms it uses, but we do know that its operating principles are close to the word embedding.
The word embedding approach
Choosing the right word analysis method is often essential for successful content optimization. Word frequency representation is the most commonly used. However, this representation has the defect of capturing little information on the context of the word(s) and their relationship to each other.
Since the late 2000s, techniques relying on artificial neural networks have become established (deep learning). This is the Word embedding.
Within a corpus, a word is not used in any way, it is contextualized: this in relation to the other words. This approach makes it possible to represent a word by a vector as a function of its position in the corpus. Word Embedding methods build a context window for each word. We use context vectors to represent each word through a matrix. Words with similar meanings are located near the terms of the vector space.
How to optimize RankBrain?
According to Google, there is no way to optimize RankBrain. RankBrain has no noticeable impact on search results, as its primary objective is to deal with queries that lack relevance.
The fact is that it is impossible for an SEO to ignore the existence of the algorithm in your SEO actions. So, what do we do ?
Expand your list of keywords beyond synonyms and co-occurrences:
No longer create pages or content suitable for a single keyword.
- Group your targeted keywords, their variations and related searches
- Use additional lexis that appear to have the same semantic concept
Focus on creating high value-added content
Be talkative, try to cover all aspects of the subject, and answer as many questions as you can.
The goal of Google and RankBrain in particular is to provide users with the most useful and relevant results. If you share this same goal, you are more likely to succeed.
Optimize your content for your audience and not for search engines
Be natural: although this statement may seem trivial, it is particularly true in the context of deep learning. Remember that the algorithm learns from human behaviour.
If people appreciate your content and consider it relevant, Rankbrain will naturally consider it relevant as well.
The world of SEO is constantly changing. The emergence of “deep learning” and word embedding have important consequences.
The RankBrain system is constantly improving through machine learning: its goal is to provide the best content to the Internet user. All you have to do is make sure your content is as relevant and complete as possible.