David Seetapun

Title: David Seetapun: Innovator in Speech Model Bias Reduction

Introduction

David Seetapun is an accomplished inventor based in Berkeley, CA. He has made significant contributions to the field of speech technology, particularly in addressing bias in production speech models. His innovative work has the potential to enhance the fairness and accuracy of speech recognition systems.

Latest Patents

David Seetapun holds a patent for "Systems and methods for principled bias reduction in production speech models." This patent describes systems and methods designed to identify and mitigate sources of bias in end-to-end speech models. The invention utilizes a recurrent neural network architecture with two 2D-convolutional input layers, followed by multiple bidirectional recurrent layers and a fully connected layer before a softmax layer. The network is trained end-to-end using the CTC loss function to predict sequences of characters from log spectrograms of audio. By optimizing recurrent layers and incorporating alignment information, the model aims to reduce unwanted bias that may arise from using purely forward-only recurrences.

Career Highlights

David is currently employed at Baidu USA LLC, where he continues to develop innovative solutions in speech technology. His work focuses on improving the reliability and fairness of speech models, which is crucial in today's diverse technological landscape.

Collaborations

David has collaborated with notable colleagues, including Eric Dean Battenberg and Rewon Child. These partnerships have fostered a creative environment that encourages the development of cutting-edge technologies in speech processing.

Conclusion

David Seetapun's contributions to the field of speech technology, particularly in bias reduction, highlight his role as a leading inventor. His innovative patent and ongoing work at Baidu USA LLC demonstrate his commitment to advancing the capabilities of speech recognition systems.