Inventor
RUOMING PANG
Inventor
US
Overview
Patents· 32 as inventor
Public. Assignee & technology derived from the patent record. Each patent links to its LexDana page.
G06Nprimary
G06NComputing arrangements based on specific computational models47%
G10LSpeech analysis techniques or speech synthesis39%
G06FElectric digital data processing13%
G06KGraphical data reading2%
PatentTitleYear
12,596,925Single-stage model training for neural architecture search202612,482,455Systems and methods for training dual-mode machine-learned speech recognition models202512,444,408Two-pass end to end speech recognition202512,373,666Convolution-augmented transformer models202512,361,927Emitting word timings with end-to-end models202512,293,276Neural architecture search with factorized hierarchical search space202512,190,869Optimizing inference performance for conformer202512,183,322Language agnostic multilingual end-to-end streaming on-device ASR system202412,175,963Synthesis of speech from text in a voice of a target speaker using neural networks202412,154,581Cascaded encoders for simplified streaming and non-streaming ASR202412,148,444Synthesizing speech from text using neural networks202412,131,244Hardware-optimized neural architecture search202412,118,988Transducer-based streaming deliberation for cascaded encoders202412,112,198Asynchronous distributed data flow for machine learning workloads202412,094,453Fast emit low-latency streaming ASR with sequence-level emission regularization utilizing forward and backward probabilities between nodes of an alignment lattice202412,079,703Convolution-augmented transformer models202412,073,824Two-pass end to end speech recognition202412,051,404Efficient streaming non-recurrent on-device end-to-end model202412,027,154Emitting word timings with end-to-end models202412,027,158Deliberation model-based two-pass end-to-end speech recognition202412,027,151Unsupervised learning of disentangled speech content and style representation202411,928,574Neural architecture search with factorized hierarchical search space202411,908,461Deliberation model-based two-pass end-to-end speech recognition202411,848,002Synthesis of speech from text in a voice of a target speaker using neural networks202311,804,212Streaming automatic speech recognition with non-streaming model distillation202311,715,458Efficient streaming non-recurrent on-device end-to-end model202311,594,212Attention-based joint acoustic and text on-device end-to-end model202311,580,956Emitting word timings with end-to-end models202311,556,381Asynchronous distributed data flow for machine learning workloads202311,545,142Using context information with end-to-end models for speech recognition202311,531,861Neural architecture search with factorized hierarchical search space202211,488,575Synthesis of speech from text in a voice of a target speaker using neural networks2022Career & activity timeline
Assembled from patent assignee/location history and declaration credentials.
2022–2026Patents assigned to GOOGLE LLC (32)· patent assignee history
2022–2026Based in US· patent location history
Inventor activity
Assignee and location history derived from the patent record.
Assignees
| Assignee | Patents | Span |
|---|---|---|
| GOOGLE LLC | 32 | 2022–2026 |
Locations
| Location | Patents | Span |
|---|---|---|
| US | 32 | 2022–2026 |