Written by Anastasia Kurmakaeva
Table of contents
A few months ago we were talking about Yandex’s AI algorithm, and how it’s been evolving over the past years since the big Palekh update in 2016. That update brought something huge to the table: the search engine’s algorithm was now using artificial intelligence neural network technology, in order to return results much more aligned with the user search intent.
Just one week before Christmas, and two weeks before the end of the year, Yandex published a detailed post on its corporate blog to let us all know that the new update had been rolled out, following in the steps of its most immediate predecessors: Palekh, Korolyov, and Andromeda. Now it’s the turn of “Vega”.
What’s new in Vega?
Just as they explain it in their blog post (which, by the way, is also available in English), Vega comes with over 1,500 new improvements made to the Russian search engine’s algorithm. The gist of this update can be summed up in four main aspects:
Palekh was the first step in a long journey, aiming to adapt its search results to the meaning of the search query with an increased accuracy, paying more attention to the user intent, instead of simple keywords. With Vega, neural networks are used at the build stage of the search database. This way, the algorithm identifies similar documents and groups them into clusters, which are created based on their meaning, long before they are requested. Then, when a user enters a search query in Yandex, the search engine doesn’t need to crawl the entirety of its database searching for an answer, but it goes straight to the cluster (or clusters), corresponding to the meaning of the input query. Thanks to this swift operation, the resources required to process the information are considerably fewer, while the results returned by Yandex are of much higher quality.
Building on the previous point, the fewer resources the search engine requires to return more accurate results to the user’s request, the faster it works. Not only that: another new feature in Vega consists of loading the results before a user finishes typing their query, something which in Yandex they call pre-rendering. The algorithm “guesses” from the first terms entered into the search bar what the full query is going to be, and displays results as soon as the user hits “Search”. This gif illustrates exactly how it works:
It also comes with instant search suggestions, which, as a whole, reminds us of the now-defunct Google Instant.
On the other hand, Yandex also developed in 2017 its own version of Google’s AMP or Baidu’s MIP, the so-called turbo-pages, which can be used to generate lighter versions of web pages, specially designed for mobile devices and slower Internet connection. In 2019 this technology has been improved. The new documentation is already available for developers and website owners alike, although, for the time being, only in the Russian language.
In my previous post (linked at the top) I had briefly mentioned this feature, which introduced the possibility of getting answers to questions from verified experts on a wide variety of topics. Yandex relies on all sorts of professionals: historians, scientists, linguists, designers, sportspeople, engineers, political scientists… Now, these people are letting the search engine know which pages respond more accurately to more complex queries (and sometimes, not even that complex). These “advising experts”, after having undergone a thorough screening process to prove their professional competence in their field, evaluate the various results shown for specific queries. This factor has considerably more weigh than other ranking signals used to teach the algorithm to be able to determine how far up or how far down should a page rank in the results.
Although Yandex was already using location to determine where a search query was made from, with Vega local searches have been further refined, to the point of knowing the neighbourhood, street, and even building number from which they are coming. So, if you’re looking for a plumber near you, and it turns out there’s one in the building right next to yours, this person’s contact will appear as the first result in Yandex. This takes us to another new feature introduced this year: Yandex Services. It helps you to find specialists advertising their services around you. Let’s say you need a private tutor for your children, and there’s a person living down the street or in the building next to yours who is registered as such in Yandex. You’ll be able to contact them and hire them to help your kids with their school work.
In 2019 we also saw the birth of the Yandex Neighbourhood platform in all cities of Russia. This social network connects you to your neighbours, providing its users with chat, message boards, announcements, polls, event posting features… You can ask people living in your quarter any daily-life related question, you can also organise a community meeting to deal with issues concerning your building, ask about a possible job in the vicinity, survey your neighbours about what is being built in X street or what store is going to open next door, etc. This social platform has been been warmly welcomed by Yandex users. However, to see the content being posted on it you must have a registered account.