Live preview - Main developer and maintainer of Kaldi: Daniel Povey was invited to participate in the INTERSPEECH2020 Magic Data live session and will interact with the audience online

BEIJING, Oct. 23, 2020 /PRNewswire/ -- At the INTERSPEECH2020 industrial forum, Magic Data as the platinum sponsor of the conference, will hold a live session at the virtual booth from 20:15 - 20:45 on Oct 28^th. The live broadcast theme is: Data sets your model --Which data strategy should be adopted to achieve better performance?

Three speech processing experts will participate in the live discussion, they are:

Daniel Povey, main developer and maintainer of Kaldi, internationally renowned professor in the field of Speech Recognition and AI, technical consultant of Magic Data, and the chief speech scientist of Xiaomi Inc.

Xuelu Zhang, linguistic/phonetic consyltant of Magic Data, research & development manager at Magic Data, PhD in Experimental Phonetics, University of Strasbourg (France). Her research focused on recognition of prosody components, automatic detection of specific linguistic phenomena and TTS.

Gaofeng Cheng, researcher of The Institute of Acoustics of the Chinese Academy of Sciences, His main research is in the field of speech recognition technology and intelligent speech platform, and he has repeatedly led his team to obtain good results in domestic speech technology evaluation. The relevant research and development achievements have been adopted by the international famous speech recognition software Kaldi and applied to practical application.

In the Magic Data live session, three guests will have an in-depth discussion on the "Data sets your model --Which data strategy should be adopted to achieve better performance?"

Discussion topics include:

How can we minimize the quantity of data that a model needs for training, and what attributes should these data have?

What kind of data we need as base in data augmentation, and what are the limits of data augmentation? 

How to exploit raw data that we collected for model training? 

What kind of data we need to pretrain a model or to fine-tune a pretrained model?

In order to give everyone an opportunity to communicate with Daniel Povey, an interactive session will be arranged at the end of the live broadcast. If you are interested in Daniel Povey's recent research or have any questions about Kaldi, you can ask Daniel Povey during the live broadcast. At that time, everyone is welcome to ask questions.

Live time:
Beijing time: Oct. 28^th, 2020  20:15 - 20:45

Virtual booth and Live session entrance:
https://zoom.com.cn/j/63391508404

About Magic Data

Beijing Magic Data Technology Company Limited (Magic Data) is one of the artificial intelligence data service providers. We have data processing centers within mainland China and a wholly owned subsidiary in Hong Kong.

Magic Data provides professional data services for enterprises and scientific research institutions engaged in the research and development and application of AI fields such as speech recognition, TTS and natural language understanding. In order to combine data with precise acoustic artificial intelligence, inspire enterprise technological wisdom, and lead innovative development ideas, Magic Data proposes a new brand slogan: Data set your mind. We hope to promote the learning, query ability and thinking method of AI through new technology and data solutions.

Magic Data has a wealth of industry experience, our dataset products and services are prospective, high-quality, multi-scenarios, and diverse. Currently, Magic Data has rich data processing experience, 50+ multilingual processing capabilities, and dozens partners in the Internet and technology industries all over the world.

Magic Data has reserved 100,000+hours of datasets series, it available for immediate application in development in various commercial domains, also can be used to train AI models immediately  The dataset products developed by Magic Data cover scenarios include smart home, smart customer service, smart assistant, smart car, smart education, and machine translation.

SOURCE Magic Data Technology