“Yandex” on June 8 introduced a new version of its browser with a personalized feed of recommendations “Zen”. This ribbon will be visible to users “Yandex.Browser” at the bottom of the search — for both mobile and desktop versions, users can opt-out of receiving access to your personal tape. In fact, “Zen” – like search, explains the head of service Victor Lambert. However, if the search is looking for specific information upon user request, the “Zen” allows you to offer the person new and interesting information for him, said Lambert.
A recommendation service is built into the browser will keep users in the browser calculates the representative of “Yandex”. In a test version of the Zen launched in 2015. During this time the number of sources to which a user contacts a week, increased from 2.7 to 4.4, and the time he spends in the browser has increased four times — up to 20 minutes, said Lambert. In addition, the integration of the service “Zen” will allow the company “to further monetize the browser at the expense of native advertising, which will appear in the film”, says the representative of “Yandex”.
The company also expects that the emergence of the recommendation service will attract new users browser (now Yandex.Browser,” according to the company, used on 30 million devices every day). From accurate predictions in “Yandex” refrain. “Zen” has already been translated into 15 languages and localized in 24 countries. “Some of them have signed contracts with mobile manufacturers about how to preinstall Yandex.Browser”, for example, in India, said Lambert. He refused to disclose what manufacturers are talking about.
To determine user preferences “Zen” uses several data types — the ones that the user is looking for in the browser (anonymized) and how he evaluates offered to him in the “Zen” article. In the basis of the recommendations lies with the service “Disco” (it is also used in “Yandex.Music” and “Yandex.Radio”). For accuracy of recommendations in the “Zen” used artificial intelligence technologies, in particular computer vision and processing of an artificial language.