FSB learns to recognize Caucasians by voice
The FSB of Russia announced a tender for “Creating an automated atlas of national languages of the Russian Federation” (R & D “D-2010-08-4.3”) with an initial contract price of 24 million rubles. and for a period of 29 months ( tactical and technical tasks in PDF ). The order is placed on behalf of part 68240, whose belonging to the FSB is known for analyzing information from open sources .
As part of the tender, it is required to develop a handbook that can become the basis for an automatic system capable of reliably recognizing the speaker's language. For example, using such a system, it is possible to quickly identify conversations in the Caucasian languages among all cellular negotiations in Moscow (provided that they are simultaneously heard through the switches of cellular operators).
The first languages for which there should be “a study of the peculiarities of the spoken language of the informant speakers” are referred to in the TTH six languages: Avarian, Ingush, Kabardino-Circassian, Karachay, Balkarian, Dargin. For each language, there should be at least 20 informants with different speech recording channels: a microphone, a telephone, etc., at least 10 recording sessions for each channel for more than 40 seconds. Then you need to analyze the sound recordings, then draw up linguistic language passports.
As part of the tender, it is required to develop a handbook that can become the basis for an automatic system capable of reliably recognizing the speaker's language. For example, using such a system, it is possible to quickly identify conversations in the Caucasian languages among all cellular negotiations in Moscow (provided that they are simultaneously heard through the switches of cellular operators).
The first languages for which there should be “a study of the peculiarities of the spoken language of the informant speakers” are referred to in the TTH six languages: Avarian, Ingush, Kabardino-Circassian, Karachay, Balkarian, Dargin. For each language, there should be at least 20 informants with different speech recording channels: a microphone, a telephone, etc., at least 10 recording sessions for each channel for more than 40 seconds. Then you need to analyze the sound recordings, then draw up linguistic language passports.
')
Source: https://habr.com/ru/post/101389/
All Articles