Yandex presented new range algorithm “Korolev” on August 22, which is based on the neuronet principles. According to the company now the search will get cleverer and will be able to answer the users’ requests not only based on key words but on the meaning.
As a matter of fact, “Korolev” is an enhanced version of the previous algorithm “Palekh”. The previous version has learnt to understand long requests which were rare because of its length and inaccuracy. However, Palekh could only identify the headings and perform the analysis of just 150 thousand documents, when the new algorithm can analyze 200 thousand documents and not only the headings but the whole text of the page.
The developers claims that “Korolev” lets you switch into another level of understanding the meaning of the texts. When a user makes his request, a robot should understand which page fits best this request to satisfy the user’s expectations. For this purposes “Palekh” used to change the requests and headlines into some particular groups of numbers – semantic vectors. These vectors can be compared with each other and the more similar the vectors are, the closer the request and the heading.
“Korolev” in its turn works not only with headlines but with the whole page. However, as this procedure demands a lot of computing resources, it happens not in real time mode, but on the stage of page indexing in advance. That’s why when the user makes a request he gets a ready outcome made in the earlier ranging stage.
It seems like there won’t be any significant differences in search results for commercial projects, so SEO specialists can take a sign of relief. The search results of high frequency commercial requests in most cases is “preserved” and leaders of the field and large internet shops head this list. “Korolev” will have a greater impact on informational projects: blogs, mass media, online magazines and thematic resource portals.