History of Yandex

Yandex launched the first conversational AI assistant that is not restricted to a set of predefined scenarios. Called Alice, it integrates multiple services to help users efficiently accomplish all kinds of tasks — find information online, plan a route, play music or get a weather forecast, among many other things. Using deep neural networks trained on massive datasets, Alice provides a user experience much like interacting with another person. It has superior Russian language skills, a distinct personality with a sense of humor, and an ability to understand incomplete phrases and questions.

The company open-sourced CatBoost, a new machine learning library based on gradient boosting. CatBoost is the successor to the machine learning algorithm MatrixNet and has numerous advantages over its predecessor: its predictions are more precise, it is more resistant to overfitting and, most importantly, it supports non-numeric features — dog breeds or cloud types, for example — that is, it can learn from different forms of data that have not been preprocessed and converted into digital form.

Yandex developed and implemented a new search algorithm, Korolyov. Like its predecessor Palekh, it is based on neural networks and searches by meaning rather than by keywords. But while Palekh only looked at headlines, Korolyov examines the entirety of web pages. It also takes into account the meaning of other queries that brought users to a web page. The update improves how Yandex handles infrequent and complex queries and delivers superior search results for its users.

The company began testing its driverless automobile technology. Two autonomous car prototypes, equipped with sensors as well as software that processes information about the surroundings and calculates the car’s trajectory, successfully found their way along their given routes, recognising and avoiding obstacles on the way.

Yandex.Taxi and Uber agreed to combine their businesses in Russia, Azerbaijan, Armenia, Belarus, Georgia and Kazakhstan. As part of the transaction, Uber was to invest $225 million with Yandex investing $100 million in cash in the new company at closing, valuing it at $3.725 billion on a post-money basis.

Yandex.Translate added a neural machine translation model to the statistical models it had been using before to implement a hybrid machine translation system. Turbo pages became available in Yandex.Search, Yandex.Zen and Yandex.News, speeding site loading on mobile devices. Yandex.Weather added weather maps covering the whole world.

The company also introduced a number of new services and products for business, including a multifunctional voice-controlled solution for connected cars, Yandex.Auto, and a workplace collaboration platform, Yandex.Connect.