Rutube Upgrades Recommendation System with AI that Classifies Video Topics with 93% Accuracy

Rutube has shifted to a new video classification based on the international taxonomy of the Interactive Advertising Bureau (IAB), which distributes digital content by topics and categories. The platform's developers created their own machine learning model that automatically determines video topics with over 93% accuracy. The new system is already operational in the service's advertising algorithms, helping to more precisely identify which videos should feature integrations. The algorithm analyzes multiple parameters: visual content, audio track, title, and description, then assigns the video to categories from the IAB structure. The model was trained on the video platform's large-scale data. The new classification includes 30 main topics and a set of expanded subcategories, increasing the number of available categories tenfold; for example, the 'Auto' section now has over 16 subtopics including repair, motorcycles, and electric vehicles. This new classification helps advertisers fine-tune targeting and simplifies analysis of campaign results. According to Rutube's product director Evgeny Ilchuk, tags will let users find needed videos faster even if the author didn't specify keywords, and in the future the classification should make recommendations even more relevant.