Yandex SpeechKit speaks Uzbek – Kommersant

The Russian company Yandex announced that Yandex SpeechKit added the ability to recognize and synthesize speech in the Uzbek language. Yandex SpeechKit will help in automating call centers and speech analytics, as well as in creating voice assistants.

The training took place on the pre-recording of the voice of a real speaker, in total, for training the model, the developers created a dataset with thousands of hours of audio and their transcripts. The neural network can synthesize not only speech, but also a phonemic record, which indicates the features of the sound in the word – this will help when pronouncing complex surnames and borrowed expressions.

Yandex SpeechKit also takes into account the peculiarities of the Uzbek language, for example, the sound of the letter “X”. According to Grigory Atrepyev, director of products at YandexCloud, “the volume of the speech technology market in Uzbekistan can reach up to 395 million minutes of processed speech per year.” Uzbek is the 16th language that the Yandex SpeechKit tool works with.

Evgeny Fedunenko