Why small and why a step?

[sumaiyakhatun27]

In other words, Yandex is learning to understand what content satisfies its audience better. The Palekh algorithm currently uses only page titles to process long-tail queries. For example, to answer the question "Pelevin's book chat", the leader of the Runet search market analyzes only the titles of web documents. The robot finds pages that are relevant to its data, gives them to users and analyzes audience satisfaction using behavioral metrics.


It accumulates data and improves the quality of the output. This is how machine learning occurs in a simplified form. Yandex still has a lot to learn: it has not yet found the correct answer to a human query. switzerland email list Kolibri analyzes the content of documents in their entirety, while Palekh is limited to headings for now.


In a post on Habr, Alexander Safronov directly writes that teaching a machine to understand documents in their entirety is a non-trivial task. Its solution requires time and resources. But Yandex wants to remain the leader of the RuNet search market. That's why Palekh will soon analyze entire documents. This will help it to better determine user needs and their satisfaction. How to work in the new reality The same as before.