New algorithm – special ranking signal
Google uses three signals in the system that help people to find the most meaningful results. The first two that have been created for simple search are content and links. The third valuable signal to be mentioned is the tool that searches for the “best fit” result that is unknown to Google.
Google processes about 15% of new search demands daily which no one has ever searched for before. You probably wonder how the number of unknown terms can be so high. We all describe places, people, and objects in different ways, sometimes unique ways. For example, users put different questions while looking for the same thing. This explains the big number of new requests Google has to deal with. There will probably be even more new queries when people start using the voice search option.
This is why the idea of creating a special processing algorithm has arisen. The “mechanism” has to offer the best matching answer to the written demand even though the system has never had a similar request before. At the beginning of utilizing RankBrain, it covered only about 15 percent of all requests. RankBrain covers almost all queries that Google considers to be new ones.
However, this innovative tool is helpful only if Google does not recognize the inquiry. So it serves as an assistant to Google when the system is unsure about user requests. When Google provides instant answers that satisfy a users’ request, the tool is powerless.
Does Google have a query set?
In 2013, when it was presented new google update – Hummingbird algorithm, it allowed users to find the best matching results by using on and off page factors to understand the relationship between people, places, and objects. The algorithm also allowed users to be transferred to more suitable pages rather than main or home pages. Therefore, writers and developers could use more convenient language instead of forced keywords.
Before, Google knew that, for example, an article about red apples was about edible fruits that come in a color known as red instead of determining an article by optimization signals such as link, anchor, text, and H1. The database tells Google that the string was “red apples,” so Google can pull back all the best results for this term.
In the same case, the system can suppose that you meant a red Apple computer and not a fruit. If Google is not sure about the correct meaning, it gives alternative results in the query set. Therefore, you will see fruit related results as well as computer-related results. This is a prime example how Google RankBrain performs and helps users to see the most suitable and useful answers.
How does the algorithm affect query results?
Google RankBrain covers queries all over the world in all languages. The algorithm works at its best when the system sees a unique query. Before RankBrain was developed it was difficult to find the necessary Internet sources that would give you information on something Google did not know.
Let’s review a good example of how RankBrain works. All states in the USA have specific water rights. If you type the keywords Las Vegas water rights, you will find the necessary results with relevant information that matches your request. However, if you type Mesquite NV water rights (a small town with a small population), Google will only give you results like mesquite trees or mesquite wood, etc. Obviously, no user has ever typed Mesquite NV water rights, therefore, Google offers results that have been typed before, such as mesquite trees or mesquite wood.
How does the whole process happen?
It is important to note that RankBrain is not a Natural Language Processor (NLP). NLP is a valuable tool that allows a computer to break down sentences of users and understand his intent. This smart tool can understand the language and how we use it. However, RankBrain cannot understand the user’s goal by language only. It requires a database of relationships between similar queries in order to provide the best matching result.
The algorithm uses databases including people, places, and things or we can say objects. The queries get broken down into word vectors. After that, world vectors receive “addresses.” So, when Google sees an unknown request, it automatically uses relationships that guess the best answer. This is how Google provides several related results, so the user can find the one that satisfies his need.
Google refines results that have been chosen by users in most cases and improves the matches between users’ intents and search results. Words “and” or “the” are not included in the analysis. However, it includes such words as “without” or “not” to improve search for negative-oriented queries.
It all seems to be a complicated scheme. To help you visualize the whole process of finding the “best fit” results, imagine content that is divided into different word sectors. Each vector has a specific address. Vectors that are located next to each other correspond to linguistic similarity. All words get assigned with a mathematical address. The matching word gets retrieved from the necessary vector based on the query. The machine learning process is constantly on. This allows the system to watch how inquiries and results match to provide more relevant result next time.
Is it possible to optimize for RankBrain?
Gary Illyes, a webmaster trends analyst at Google, says that optimizing for the algorithm is very easy. The only recommendation is to write in the natural language and avoid forced keywords that complicate searching. Therefore, users should use the language we all speak and understand.
If you have a website, ask people to read its content and tell you whether it sounds natural. If so, then you are automatically optimized for RankBrain. You probably complicate your work and lose time by trying to optimize words or phrases that no one is interested in and no one searches for.
Anyway, the request always changes, so you will really have to work hard to keep up with new demands and create matching results. RankBrain is designed to change “unknown” results and deliver results that can satisfy users. Do not waste your time trying to optimize keywords that constantly change. Still, the algorithm can be useful for some unique cases.
The conclusion is to write good content that is easy to perceive and sounds clear using words that you pronounce every day to communicate with people.