|
"Palekh" is Yandex search engine algorithm for low-frequency requests, introduced by Yandex in 2016. It is named in honor of a village in the Ivanovo region, where the coat of arms is a fire-bird.
It is with the tail of the firebird that low-frequency requests & their complexity, are associated. Here is a schematic image provided by Yandex itself.
Palekh algorithm schematic
The beak is a high-frequency request. Body - medium frequency.
The fact is that out of 280 million daily requests, 100 million are low frequency. Recently, webmasters have started to use low frequency queries, also known as long-tail keywords, more and more often, since these keywords don't have much competition and it's easier to get to the top of search results. On the flip side, low-frequency requests are rare and its impossible to guess everything.
The essence of Palekh algorithm is that the ranking began to be carried out not only by the entered search query, but also by other phrases with a similar meaning. That is, now Yandex is trying not just to find matches in the query and the webpage, but to understand the meaning of the search and find the webpages that satisfy this request.
Artificial neural networks
In addition to low frequency queries, there are many unique queries which are entered once. They are difficult to process by the search engine, because there are no detailed statistics on them. Therefore, Yandex developers decided to use artificial neural networks (ANNs). It was the ANN that formed the basis of the "Palekh" algorithm. The main feature of such networks is the ability to learn & self-correct. ANNs are created in the image of the human brain and can absorb a huge amount of information. On its basis, ANN conducts an analysis and pulls out the correct result.
Why was Palekh created?
It is important for Yandex to understand whether it satisfied the user's needs. That is, did the person find what he was looking for? When the user enters a query, goes to the first site and does not return to the search again, Yandex realizes that it has shown the correct result. In the future, Yandex will show it to other people. In the case of high-frequency requests, there is a lot of such data and, therefore, it is easy to choose the correct output. With low-frequency requests, everything is more complicated. Yandex needed a lot of time to find the optimal version of the output for a low frequency requests. And the users themselves foten could not find what they were looking for on their first try. "Palekh" algorithm was created precisely to solve this problem.
|
