Making the right choice for voice. Sathish Sastry, Head of the 3i Voice Innovation Centre at Zetes, explains the pros and cons of trained (voice recognition) and untrained (speech recognition) voice systems.

What is the difference between a trained and an untrained voice system?

Sathish Sastry: “An untrained voice system is ready to use, right out of the box, without any registering of the operator’s voice. A trained voice system first needs registering the user’s voice with the various commands.”

Is there a difference in accuracy between a trained and an untrained voice system?

Sathish Sastry: “A trained system is by default more accurate, because the user trains the voice template before he or she starts using the system. However, untrained voice systems have made a lot of progress over the years, delivering results that are equal to those of a trained voice system.”

Does this mean an untrained voice system can be as successful as a trained one?

Sathish Sastry: “It actually can. The new generation of untrained voice systems has a lot more information already built into them. When using suitable grammar for the application that the untrained system supports, it can become even better than a trained one. For a trained system, it actually makes no difference whether or not the user speaks an existing language. As trained systems do not depend on an actual language, you can easily train it with sounds that are not real words. To extend the use of an untrained system, you need to add a new language set. This may be necessary to cover regional accents also.”

What are the main drivers for choosing either a trained or an untrained voice system?

Sathish Sastry: “It all comes down to the environment the system is required for. A trained voice system is ideal for a stable environment, with a constant  team of users. In reality however, quite a number of organizations are facing a high turnaround of employees who may be active for only a couple of days. As the registering of a new profile on a trained voice system can take up to half an hour then in this case the use of an untrained voice system with guided dialogue could be more efficient.”

 

Following a rebranding exercise, the Zetes’ logistics execution solution using voice technology is now called ZetesMedea Voice (formerly 3i Voice). All features and functionalities remain valid.