A chatbot developed by Sber for the Moscow Metro handles 88% of inquiries without a human operator



Speech Technology Center (part of the Sber ecosystem), together with the Moskovsky Bank of Sberbank and the team of Moscow Metro, introduced a chatbot for passengers. During testing in a focus group, the virtual text assistant automatically processed 75% of inquiries, without transferring them to human operators. From 17 November, the virtual assistant is running in commercial operations, where it achieved 88% automation. The bot is available through the Moscow Metro app.

Roman Latypov, First Deputy Head of the Metro for Strategic Development and Customer Management: 'Our metrobot is becoming one of the most technologically advanced and modern customer information services. The virtual assistant demonstrated excellent performance even in the first days of its operation, as only 12% of cases required help from a human operator to provide the response to the inquiries (such inquiries usually were too long or included confusing wording). The chatbot can handle implemented dialogue scenarios for more than 17,000 questions. Alexandra (this name was selected by our passengerscan inform about the fares for the rides and opening hours of stations and transfer points not only for the metro, but also for the Moscow Central Circle (MCC), Moscow Central Diameters (MCD), and this is only the start, as we plan to expand this list in the future'.

Gartner analysts forecast that, by 2024, nearly 70% of routine operations can be automated using artificial intelligence, including the chatbots. The virtual assistant developed by Speech Technology Center for the Moscow Metro demonstrates 88% automation even today.

The team of the Speech Technology Center created the chatbot using ChatNavigator, its own product that provides an omnichannel dialogue platform for creating intelligent virtual assistants. The chatbot implements a specially designed universal connector (API), which allows communicating with passengers across various channels without assistance from the operator. While the basic communication scenario remains the same for all channels, the assistant takes into account the communication-related specifics of each channel.

The most popular topics of inquiries to the bot during the testing in the focus group included the fare policy and fare payment methods, fare privileges, metro services, and general questions about the quality of services.

Dmitriy Dyrmovskiy, CEO of the Speech Technology Center: 'A special feature of this project was the fact that all components of the chatbot and platform are hosted in the cloud—this is one of the first completely cloud-based solutions on the market. To ensure further development of the bot and more accurate understanding of inquiries, we use a hybrid classification based on a combination of rules and machine learning. This will enable the virtual assistant to communicate on new topics and better handle the existing ones to ensure continuous improvement'.

Vyacheslav Tsybulnikov, Vice President of Sberbank, Chairman of Moskovsky Bank: 'Sberbank offers Muscovites and visitors of Moscow the most advanced digital services in financial and many other areas of everyday life. We are confident that the chatbot developed with the participation of Speech Technology Center, our subsidiary company, will make communication with passengers more efficient and user-friendly not only for the metro but, in the future, other types of public transport, as well'.

Speech Technology Center (part of the Sberbank ecosystem) is a global developer of intelligent speech and face recognition technologies and an expert resource in AI and machine learning with 30 years of experience in the field. Speech Technology Center is one of the few companies in the world that creates and develops both biometric modalities: face and voice. The voice-falsification detection and speech-recognition solutions developed by Speech Technology Center occupy leading positions in global rankings by NIST, ASVspoof Challenge, and VOiCES. CHiME 2020 named Speech Technology Center’s speech recognition technology the best in the world. Speech Technology Center has used artificial intelligence to implement more than 5,000 projects in 70 countries.
Media contact: pr@speechpro.com