New Mnemonic recommendation algorithm launched in Kinopoisk app

A new recommendation algorithm called Mnemonic has been launched in the “What to watch” section of the Kinopoisk app. The main task of this system is to create a personalized video feed of trailers for each user, taking into account their interests across various platforms. Unlike previous systems, Mnemonic analyzes viewer behavior directly within the feed in real time. This is reported by Ixbt.com reports .

Positive signals for the algorithm include watching a video for more than 30 seconds, visiting a movie card, clicking the “Watch” button, or adding it to the “Watch later” list. The more actively a user interacts with the content, the more accurate the recommendations become. This mechanism is similar to the operating principles of social networks that provide short-form videos.

The system “remembers” not only the viewing history within Kinopoisk but also interests in content on other streaming services such as Netflix, Apple TV, Amediateka, and Start. The Kinopoisk encyclopedia, including the Kinopoisk Pro index which evaluates series popularity, serves as the database. The Mnemonic technology is based on the CatBoost machine learning model, which can recognize over 1,000 behavioral signals.

Simultaneously with the algorithm's launch, the section interface was also updated. Users can now rate content directly from the trailer card, mark it as “Watched,” and view recommendations from friends. The name Mnemonic refers to the ability to build associations like human memory, serving to filter data more accurately.