-
公开(公告)号：US11587569B2
公开(公告)日：2023-02-21
申请号：US15931788
申请日：2020-05-14
Applicant: MICROSOFT TECHNOLOGY LICENSING, LLC
Inventor： Guoli Ye , Yan Huang , Wenning Wei , Lei He , Eva Sharma , Jian Wu , Yao Tian , Edward C. Lin , Yifan Gong , Rui Zhao , Jinyu Li , William Maxwell Gale
Abstract: Systems, methods, and devices are provided for generating and using text-to-speech (TTS) data for improved speech recognition models. A main model is trained with keyword independent baseline training data. In some instances, acoustic and language model sub-components of the main model are modified with new TTS training data. In some instances, the new TTS training is obtained from a multi-speaker neural TTS system for a keyword that is underrepresented in the baseline training data. In some instances, the new TTS training data is used for pronunciation learning and normalization of keyword dependent confidence scores in keyword spotting (KWS) applications. In some instances, the new TTS training data is used for rapid speaker adaptation in speech recognition models.
-
公开(公告)号：US10834191B2
公开(公告)日：2020-11-10
申请号：US15898105
申请日：2018-02-15
Applicant: MICROSOFT TECHNOLOGY LICENSING, LLC
Inventor： Nicolaas Deodorus Peelen , Wang Hui , Jun Tang , Sridhar Srinivasan , Mingqiang Xu , Yan Huang
IPC: H04L29/08 , H04L12/801
Abstract: In various embodiments, methods and systems for enhanced access to storage data based on a collaboration data proxy system are provided. A plurality of metadata tables on one or more peer nodes are referenced for data corresponding to a data request of a requesting node. The metadata tables indicate availability of chunks of data in the one or more peer nodes. A determination is made that the data corresponding to the data request is downloadable from the one or more node; the determination is based on the metadata tables. A download operation configuration instance is generated for a data request of a requesting node. The download operation configuration instance comprises configuration settings for downloading data corresponding to the data request from the one or more peer nodes. The chunk of data is downloaded from the corresponding one or more peer nodes where the chunk is located, using the configuration settings.
-
Patent Agency Ranking
公开(公告)号：US09906597B2
公开(公告)日：2018-02-27
申请号：US14680757
申请日：2015-04-07
Applicant: MICROSOFT TECHNOLOGY LICENSING, LLC
Inventor： Nicolaas Deodorus Peelen , Wang Hui , Jun Tang , Sridhar Srinivasan , Mingqiang Xu , Yan Huang
IPC: H04L29/08 , H04L12/801
CPC classification number: H04L67/1097 , H04L47/12 , H04L67/06 , H04L67/104 , H04L67/28
Abstract: In various embodiments, methods and systems for enhanced access to storage data based on a collaboration data proxy system are provided. A plurality of metadata tables on one or more peer nodes are referenced for data corresponding to a data request of a requesting node. The metadata tables indicate availability of chunks of data in the one or more peer nodes. A determination is made that the data corresponding to the data request is downloadable from the one or more node; the determination is based on the metadata tables. A download operation configuration instance is generated for a data request of a requesting node. The download operation configuration instance comprises configuration settings for downloading data corresponding to the data request from the one or more peer nodes. The chunk of data is downloaded from the corresponding one or more peer nodes where the chunk is located, using the configuration settings.
-
公开(公告)号：US11861427B2
公开(公告)日：2024-01-02
申请号：US16671672
申请日：2019-11-01
Applicant: Microsoft Technology Licensing, LLC
Inventor： Jason Michael Anderson , Soumya Desai , Vrijesh Kothari , Marc Edward Mercuri , Yan Huang
CPC classification number: G06F9/547 , G06F16/2379 , G06F16/27 , G06F16/953 , H04L9/3239 , H04L9/50
Abstract: The disclosed technology is generally directed to blockchain technology. In one example of the technology, a first transaction node of a hosted permissioned blockchain network is provisioned for a first consortium member of the hosted permissioned blockchain network. A shared pool of validator nodes of the hosted permissioned blockchain network is provisioned. The shared pool of validator nodes includes at least one validator node. The shared pool of validator nodes is shared among the plurality of consortium members. The validator nodes of the shared pool of validator nodes are configured for blockchain transaction validation based on a BFT consensus protocol. A second transaction node of the hosted permissioned blockchain network is provisioned for a second consortium member of the hosted permissioned blockchain network. Each transaction node of the hosted permissioned blockchain network is separate from each validator node of the hosted permissioned blockchain network.
-
公开(公告)号：US20210304769A1
公开(公告)日：2021-09-30
申请号：US15931788
申请日：2020-05-14
Applicant: MICROSOFT TECHNOLOGY LICENSING, LLC
Inventor： Guoli Ye , Yan Huang , Wenning Wei , Lei He , Eva Sharma , Jian Wu , Yao Tian , Edward C. Lin , Yifan Gong , Rui Zhao , Jinyu Li , William Maxwell Gale
Abstract: Systems, methods, and devices are provided for generating and using text-to-speech (TTS) data for improved speech recognition models. A main model is trained with keyword independent baseline training data. In some instances, acoustic and language model sub-components of the main model are modified with new TTS training data. In some instances, the new TTS training is obtained from a multi-speaker neural TTS system for a keyword that is underrepresented in the baseline training data. In some instances, the new TTS training data is used for pronunciation learning and normalization of keyword dependent confidence scores in keyword spotting (KWS) applications. In some instances, the new TTS training data is used for rapid speaker adaptation in speech recognition models.
-
公开(公告)号：US10235994B2
公开(公告)日：2019-03-19
申请号：US15199346
申请日：2016-06-30
Applicant: MICROSOFT TECHNOLOGY LICENSING, LLC
Inventor： Yan Huang , Chaojun Liu , Kshitiz Kumar , Kaustubh Prakash Kalgaonkar , Yifan Gong
IPC: G06N3/04 , G10L15/02 , G10L15/06 , G10L15/16 , G10L15/28 , G10L15/065 , G10L15/183
Abstract: The technology described herein uses a modular model to process speech. A deep learning based acoustic model comprises a stack of different types of neural network layers. The sub-modules of a deep learning based acoustic model can be used to represent distinct non-phonetic acoustic factors, such as accent origins (e.g. native, non-native), speech channels (e.g. mobile, bluetooth, desktop etc.), speech application scenario (e.g. voice search, short message dictation etc.), and speaker variation (e.g. individual speakers or clustered speakers), etc. The technology described herein uses certain sub-modules in a first context and a second group of sub-modules in a second context.
-
公开(公告)号：US12205596B2
公开(公告)日：2025-01-21
申请号：US18108316
申请日：2023-02-10
Applicant: Microsoft Technology Licensing, LLC
Inventor： Guoli Ye , Yan Huang , Wenning Wei , Lei He , Eva Sharma , Jian Wu , Yao Tian , Edward C. Lin , Yifan Gong , Rui Zhao , Jinyu Li , William Maxwell Gale
Abstract: Systems, methods, and devices are provided for generating and using text-to-speech (TTS) data for improved speech recognition models. A main model is trained with keyword independent baseline training data. In some instances, acoustic and language model sub-components of the main model are modified with new TTS training data. In some instances, the new TTS training is obtained from a multi-speaker neural TTS system for a keyword that is underrepresented in the baseline training data. In some instances, the new TTS training data is used for pronunciation learning and normalization of keyword dependent confidence scores in keyword spotting (KWS) applications. In some instances, the new TTS training data is used for rapid speaker adaptation in speech recognition models.
-
-
-
-
-
-
Search Results
7
records
(took 0.016 seconds)
