Azerbaijan, Baku, June 7 / Trend H. Valiyev /
Dilmanc project will put its demo-version of speech recognition systems through Internet, Dilmanc Project Manager Abulfat Fatullayev told Trend.
He said that the system can recognize 130,000 words and phrases in the Azerbaijani language. The system uses modules Acoustic and Language. Module Language is built on the shell, consisting of 150 million of word forms, and the module Acoustic - 50-hour speech. About 1,500 speakers' voices are used in the system.
The system is convenient for entering text via a microphone. Language filters are configured on the peculiarities of pronunciation that can translate spoken words into text in any editor in Windows. At present, the speed of transfer of large sentences in electronic form hits 15-16 seconds. In future, that figure will improve.
At present, specialists of the project are working to improve the system. It will allow using voice commands to manage e-mail, work with different programs, etc.