We use cookies to make sure our website better meets your expectations.
You can adjust your web browser's settings to stop accepting cookies. For further information, read our cookie policy.
IN Warsaw
Exchange Rates
Warsaw Stock Exchange - Indices
The Warsaw Voice » Special Sections » August 1, 2014
AGH University of Science and Technology
You have to be logged in to use the ReadSpeaker utility and listen to a text. It's free-of-charge. Just log in to the site or register if you are not registered user yet.
Talking to the Computer
August 1, 2014   
Article's tools:

Sarmata is a computer speech recognition system that recognizes Polish words – a rare feature in an ocean of technology based on English language. Designed by researchers from the AGH University of Science and Technology, the system recognizes a total of 1,000 commands and enables Polish users to communicate with digital devices by voice.

After fine-tuning, the system will be put to use in telecommunications and banking. Techmo, a spin-off technology company co-owned by the AGH UST, is taking care of the commercial side of the project – putting the Sarmata system on the market. The company is managed and co-owned by the system’s inventors, Bartosz Ziółko, Ph.D., Dawid Skurzok, M.Sc., Professor Mariusz Ziółko, Jakub Gałka, Ph.D., and Tomasz Jadczyk, M.Sc.

The system makes it possible to issue voice commands to a phone, tablet or computer instead of using a mouse or touchscreen. Telecommunications companies which decide to buy the system will not have to employ human telephone operators. Users will be able to ask the software for their call to be put through to the customer service department, front desk or warehouse. The system will automatically switch the caller to the appropriate line, with no need for the user to click any buttons on their cell phone.

The Sarmata system makes use of Surikate and VoicePass technology to automatically recognize the caller, which will make it possible, for example, to carry out banking operations.

Although these technologies are still being developed, the Techmo company went commercial with the system a year ago in order to actively look for customers in the IT industry. This decision will enable it to customize solutions at the implementation stage to the individual needs of a business partner.

“The Sarmata software is primarily expected to work in IVR [interactive voice response] systems, which means when we call and talk to a machine. A further application is calling an infoline or a bank to verbally make a transfer,” says Dawid Skurzok.

The company provides technical support to customers at an early stage. The AGH UST retains the intellectual property rights to the system.

The processing of speech and natural language into written text is a major challenge for modern science. There are many potential uses for the speech recognition systems being developed at the AGH UST laboratories.

Techmo provides telephone exchanges with software that makes it possible to avoid a long wait for having your call put through. The caller speaks a password out loud and the exchange puts their call through.

According to the researchers, creating a speech recognition system adapted to Polish users is a major challenge. The system must be “trained” in the right way, and to this end specific word patterns must be fed into it. To simplify the process, special computer programs are created that divide long recordings into smaller fragments, sentences or words.

Various ready-made recordings and text are needed for this type of research. Google, Microsoft and IBM have databases with hundreds of thousands of hours of recordings in English. Research into English-speaking systems began 60 years ago. In Poland, such systems began to be developed only a decade ago.

Speaker identification systems have high potential for commercial use. The university’s electronics department has developed a recognition system based on voice characteristics. It analyzes the tone of the voice as well as the biometric features of the speaker. But it is impossible to recognize the identity of an unknown person; the system requires prior exposure to an individual’s voice.

Sarmata is a promising solution with a future that offers enhanced security and user authentication. The invention has won a silver medal at an innovation fair in Taiwan and a silver medal at the IWIS invention show in Warsaw.

Karolina Olszewska
Latest articles in Special Sections
Latest news in Special Sections
Mercure - The 6 Friends Theory - Casting call
© The Warsaw Voice 2010-2018
E-mail Marketing Powered by SARE