Fun-ASR - A New Generation of Speech Recognition Models Jointly Launched by Nail and Tongyi

What is Fun-ASR

Fun-ASR is a big model of speech recognition jointly launched by Nail and Tongyi Labs. The model has been trained with massive audio data and can accurately recognize multi-industry terminology, such as Internet, technology, home decoration, etc., significantly improving the recognition accuracy. The model combines with Nail's enterprise information for inference optimization, reduces phantom problems and provides reliable transcription results.Fun-ASR supports enterprise-specific customized training, optimizes the algorithm with real enterprise speech data, and improves the recognition accuracy of exclusive words. Fun-ASR has been integrated into Nail's meeting subtitles, intelligent minutes, voice assistant and other functional modules, providing enterprises with stable, efficient and easily scalable speech recognition solutions to meet their demanding speech recognition needs.

Main functions of Fun-ASR

• Precise identification of terms: It can accurately recognize terminology from multiple industries (e.g., Internet, technology, home improvement, animal husbandry, etc.), significantly improving the recognition accuracy.
• contextual optimization: Combine inference optimization with enterprise information (e.g., address book, calendar, knowledge base, etc.) within the nail to reduce the model illusion problem and provide more reliable transcription results.
• Customized training: Support enterprises to use their own real speech data for customized training to further improve the recognition accuracy of exclusive words (e.g. brand names, project code names, etc.).
• multi-scenario application: It has been integrated into several functional modules of Nail (e.g. meeting subtitles, intelligent minutes, voice assistant, etc.), providing enterprises with stable, efficient, and easily scalable voice recognition services to meet the high requirements in different scenarios.

Core Benefits of Fun-ASR

• High-precision recognition: After massive data training, it can accurately recognize professional terms of many industries, significantly improve the recognition accuracy, and meet the high requirements for speech recognition in different industries.
• Deep customizationIt supports enterprises to customize the training of exclusive models according to their own needs, and optimize the algorithm with real voice data from enterprises to further improve the recognition accuracy of exclusive words and better adapt to enterprise-specific scenarios.
• context-sensitive: Combine the inference optimization with the enterprise information within the nail, effectively reduce the phantom problems that may occur in the model, provide more reliable and accurate transcription results, and improve the user experience.
• Continuous optimization: Based on an efficient end-to-end training architecture, it can continuously optimize with new data to keep the model advanced and accurate and adapt to changing speech recognition needs.

Who is Fun-ASR for?

• management: Efficient meeting minutes and smart summary features are needed to facilitate quick capture of meeting points and action items.
• business unit: e.g. sales, marketing, customer service, etc., need to accurately recognize professional terminology to improve customer communication and service quality.
• Technical Team: e.g. R&D, Ops, etc., need to quickly document and understand complex technical terms in technical exchanges and meetings.
• Internet and technology industry: The need to recognize a lot of jargon and technical vocabulary enhances productivity.
• home improvement industry: Need to accurately identify material names, design terminology, etc. to improve customer communication and service.