logo

For the Inventor, By the Inventor

Bellevue, WA, United States of America

Jia Cui

USPTO Granted Patents = 10 

Average Co-Inventor Count = 5.1

ph-index = 1

Forward Citations = 5(Granted Patents)


Company Filing History:

goldMedal10 out of 1,861 
 
Tencent America LLC
 patents

Years Active: 2020-2024

Loading Chart...
10 patents (USPTO):Explore Patents
1. 11972754 - Multi-task training architecture and strategy for attention-based speech recognition system (2024-04-30),
2. 11803618 - N-best softmax smoothing for minimum bayes risk training of attention based sequence-to-sequence models (2023-10-31),
3. 11636848 - Token-wise training for attention based end-to-end speech recognition (2023-04-25),
4. 11551136 - N-best softmax smoothing for minimum bayes risk training of attention based sequence-to-sequence models (2023-01-10),
5. 11257481 - Multi-task training architecture and strategy for attention-based speech recognition system (2022-02-22),
6. 11037547 - Token-wise training for attention based end-to-end speech recognition (2021-06-15),
7. 11004443 - Multistage curriculum training framework for acoustic-to-word speech recognition (2021-05-11),
8. 10923117 - Best path change rate for unsupervised language model weight selection (2021-02-16),
9. 10861441 - Large margin training for attention-based end-to-end speech recognition (2020-12-08),
10. 10672382 - Input-feeding architecture for attention based end-to-end speech recognition (2020-06-02),

Title: Jia Cui: Innovator in Speech Recognition Technology

Introduction

Jia Cui is a prominent inventor based in Bellevue, WA (US), known for his significant contributions to the field of speech recognition technology. With a total of 10 patents to his name, he has made remarkable advancements that enhance the efficiency and accuracy of speech recognition systems.

Latest Patents

One of Jia Cui's latest patents is titled "Multi-task training architecture and strategy for attention-based speech recognition system." This patent outlines methods and apparatuses for performing sequence to sequence (Seq2Seq) speech recognition training using at least one processor. The method involves acquiring a training set that includes pairs of input data and corresponding target data, encoding the input data into a sequence of hidden states, and performing both connectionist temporal classification (CTC) model training and attention model training based on these hidden states. The process culminates in decoding the sequence of hidden states to generate target labels through independent training of the CTC and attention models.

Another notable patent is "N-best softmax smoothing for minimum bayes risk training of attention-based sequence-to-sequence models." This invention provides methods and apparatuses for analyzing sequence-to-sequence data, such as speech data or machine translation data, by employing minimum Bayes risk (MBR) training. It introduces applications of softmax smoothing to enhance the N-best generation during MBR training of the sequence-to-sequence model.

Career Highlights

Jia Cui is currently employed at Tencent America LLC, where he continues to innovate and develop cutting-edge technologies in speech recognition. His work has significantly impacted the way speech recognition systems are trained and implemented, making them more effective and reliable.

Collaborations

Jia has collaborated with notable colleagues, including Chao Weng and Dong Yu, who have contributed to his research and development efforts in the field.

Conclusion

Jia Cui's innovative work in speech recognition technology, marked by his numerous patents, showcases his dedication to advancing this critical field. His contributions are paving the way for more sophisticated and efficient speech recognition systems in the future.

This text is generated by artificial intelligence and may not be accurate.
Please report any incorrect information to support@idiyas.com
Loading…