A more advanced algorithm
Posted: Wed Dec 18, 2024 10:30 am
Palekh and Korolev certainly improved the ranking, but they still did not cope with this task well enough. Among the shortcomings of their work, the following can be highlighted:
There was no uniform accounting of word order.
The meaning of the page described a single vector of limited size.
Now each text element is transformed into a special vector that preserves its location in the text. The training principle of the new search engine algorithm is transfer learning. First, information is collected that helps solve a problem. The same information is then used to solve similar problems.
At first, YATI is "fed" with simple, not the most reliable 100% active phone number list information (from Yandex. Toloka). This is how the first stage of training takes place. Only after this, the "transformer" is given expert information, but in a smaller volume (from Yandex assessors).
The main difference from previous ranking systems: the new algorithm has learned to predict where the user will click. This additional metric will affect the search results.
Now text analysis will become more profound. The algorithm can understand the meaning of all the information on the site, which means that the role of semantic load will increase. Texts will have to be written differently, they must evolve. In the near future, LSI copywriting will develop, which has the following features:
The text is written based on keywords, but attention is paid not to the incoming keys, but to the semantic load.
The text includes not only the main keywords, but also those related to the main queries.
The text material must not only be unique, but also meaningful, useful, and exciting.
Only texts that contain expert information and provide complete, detailed information will increasingly be displayed in the top search positions.
It will be possible to improve the texts by structuring them. It is advisable to put all the keys in the headings. The texts will also improve after working out the title and description. The semantics for the pages should be collected in full.
What about Google's BERT?
A year ago, Google connected its "transformer" BERT. It analyzes search queries, their context, and not just key queries separately. It understands sentences completely. BERT, like YATI, is aimed at understanding the content of texts. But BERT cannot predict where the user will click. This is its imperfection.
About old SEO methods
Search engines are constantly improving. Many may think that the methods that were used before for optimization will no longer work. These worries are unfounded.
Optimization of H headers and title and description tags is still relevant. The ranking factors that were introduced earlier have not been cancelled, they have only been supplemented with a more qualitative analysis of texts.
Technical optimization, adding natural links, improving behavioral factors - all this should not be stopped. A good result from SEO can only be achieved with complex work.
Conclusion
Yandex search results will change soon. They are self-learning, so this will not happen right away. There is still time to refine the texts. They can still be adapted to modern search engine standards.
This is exactly what our Web-studio AXI does. Contact us, and we will rework the texts of your site so that it does not lose its positions in the search results.
There was no uniform accounting of word order.
The meaning of the page described a single vector of limited size.
Now each text element is transformed into a special vector that preserves its location in the text. The training principle of the new search engine algorithm is transfer learning. First, information is collected that helps solve a problem. The same information is then used to solve similar problems.
At first, YATI is "fed" with simple, not the most reliable 100% active phone number list information (from Yandex. Toloka). This is how the first stage of training takes place. Only after this, the "transformer" is given expert information, but in a smaller volume (from Yandex assessors).
The main difference from previous ranking systems: the new algorithm has learned to predict where the user will click. This additional metric will affect the search results.
Now text analysis will become more profound. The algorithm can understand the meaning of all the information on the site, which means that the role of semantic load will increase. Texts will have to be written differently, they must evolve. In the near future, LSI copywriting will develop, which has the following features:
The text is written based on keywords, but attention is paid not to the incoming keys, but to the semantic load.
The text includes not only the main keywords, but also those related to the main queries.
The text material must not only be unique, but also meaningful, useful, and exciting.
Only texts that contain expert information and provide complete, detailed information will increasingly be displayed in the top search positions.
It will be possible to improve the texts by structuring them. It is advisable to put all the keys in the headings. The texts will also improve after working out the title and description. The semantics for the pages should be collected in full.
What about Google's BERT?
A year ago, Google connected its "transformer" BERT. It analyzes search queries, their context, and not just key queries separately. It understands sentences completely. BERT, like YATI, is aimed at understanding the content of texts. But BERT cannot predict where the user will click. This is its imperfection.
About old SEO methods
Search engines are constantly improving. Many may think that the methods that were used before for optimization will no longer work. These worries are unfounded.
Optimization of H headers and title and description tags is still relevant. The ranking factors that were introduced earlier have not been cancelled, they have only been supplemented with a more qualitative analysis of texts.
Technical optimization, adding natural links, improving behavioral factors - all this should not be stopped. A good result from SEO can only be achieved with complex work.
Conclusion
Yandex search results will change soon. They are self-learning, so this will not happen right away. There is still time to refine the texts. They can still be adapted to modern search engine standards.
This is exactly what our Web-studio AXI does. Contact us, and we will rework the texts of your site so that it does not lose its positions in the search results.