Guangzhou, May 27, 2022 -- In recent years, with the promotion of the Belt and Road Initiative, resource-rich Xinjiang, as the core region of the Silk Road Economic Belt, has become an important fulcrum of economic growth in China's western region. With increasing integration with other ethnic groups, more and more people have come to know Xinjiang.
In Xinjiang, nearly 50% of the population is Uyghur. Uyghur is the main language spoken, and a large proportion of them do not know Chinese.
With the rapid development of AI, application scenarios continue to expand, and intelligent voice devices are everywhere, the demand for dialect speech recognition is also increasing. In order to enable Uyghur people to enjoy the convenience of work and life brought by new technologies such as artificial intelligence, big data and cloud computing, Standard & Babs launched Uyghur speech recognition service based on the deep learning platform and a large number of Uyghur vocabulary, to facilitate the exchange of business culture between Uyghur and Han, and promote the all-round development of local economy and society.
Speech recognition ability of standard Bevy language
Speech recognition is to solve the problem of making the machine understand, but it is affected by complex external factors, such as environmental noise, multi-person conversation, dialect accent, etc., will cause certain interference to the recognition results. Once the recognition is wrong, it may affect the understanding of the communication parties to the information.
Based on the self-developed deep neural network training acoustic model, and a large number of Uighur corpus data for model and system iterative tuning, the final output can be commercialized Uighur speech recognition service capability, the overall recognition speed and accuracy can meet the personalized needs of various speech interaction scenarios.
For example, in the field of intelligent customer service, intelligent call quality inspection is conducted based on the call recording between artificial agents and customers to help customer service improve service quality. In terms of the application of government affairs, it can provide intelligent conference voice transliteration scheme for public security, judicial and other institutions in Xinjiang, and intelligent real-time voice transliteration system for trial for courts, so as to smooth information communication and effectively improve the business efficiency of political and legal institutions. In the online education scene, accurately identify and analyze the oral pronunciation and expression ability of Uygur learners, so as to rapidly improve their oral ability.
Standard Bevy speech database
As we all know, various technologies based on machine learning are often inseparable from the accumulation of algorithms and data. In order to improve the accuracy of speech recognition, a large number of high-quality speech data is needed as the support of model training.
Uyghur is one of the official languages of Xinjiang Uygur Autonomous Region and is spoken by about 15 million people in China. Due to the characteristics of adhesive language, the use of abundant affixes can produce super-large words, making it more difficult to collect and label Uyghur speech than other domestic languages. As a result, Uyghur speech recognition training corpus is always scarce, bringing great difficulties to speech recognition.
In the face of the above problems, before the launch of the Uighur speech recognition service, Standard & Bey Technology has launched 800 hours of adult Uighur reading and free conversation database, more than 1000 people participated in the recording, has completed the annotation, data quality to meet the requirements of commercialization.
Uyghur reading database for adults
Database features: reading class voice
Recording environment: quiet room
Data duration: 600 hours
Number of recordings: 605
Recording corpus: universal
File format: WAV
Voice parameters: 16kHz/16bits
Recording device: Cell phone
Application field: It can be used in voice recognition scenarios such as intelligent customer service and smart home
Adult Uygur Free Conversation Database
Database features: free conversation class voice
Recording environment: quiet room
Data duration: 200 hours
Number of recordings: 450
Recording corpus: universal
File format: WAV
Voice parameters: 16kHz/16bits
Recording device: Cell phone
Application: It can be used in intelligent conference system, input method, social communication and other speech recognition scenarios
Welcome industry partners interested in the above data set to contact us ~
With the Uyghur speech recognition capability online, the current standard and Bei technology can support Chinese characters, English; The speech recognition of Cantonese and Uygur in dialects is widely used in work, life and study. In the future, on the basis of technological innovation and data services, Standard & Bei Technology will continue to create more accurate and efficient voice recognition services for the AI industry.
