公开(公告)号：EP4285358A1

公开(公告)日：2023-12-06

申请号：EP22710892.5

申请日：2022-02-28

Applicant: Google LLC

Inventor： PEDDINTI, Vijayaditya ,  RAMABHADRAN, Bhuvana ,  ROSENBERG, Andrew ,  GOLEBIEWSKI, Mateusz

IPC: G10L13/08 ,  G10L15/22 ,  G10L25/30

公开(公告)号：EP4118641A1

公开(公告)日：2023-01-18

申请号：EP21727634.4

申请日：2021-04-27

Applicant: Google LLC

Inventor： CHEN, Zhehuai ,  ROSENBERG, Andrew ,  RAMABHADRAN, Bhuvana ,  MORENO MENGIBAR, Pedro J.

IPC: G10L13/00

公开(公告)号：EP3877973A1

公开(公告)日：2021-09-15

申请号：EP19707226.7

申请日：2019-02-08

Applicant: Google LLC

Inventor： RAMABHADRAN, Bhuvana ,  MA, Min ,  MENGIBAR, Pedro J. Moreno ,  EMOND, Jesse ,  ROARK, Brian E.

IPC: G10L15/06 ,  G06F40/40

公开(公告)号：EP4476718A1

公开(公告)日：2024-12-18

申请号：EP22722960.6

申请日：2022-04-25

Applicant: GOOGLE LLC

Inventor： ROSENBERG, Andrew, M. ,  WANG, Gary ,  RAMABHADRAN, Bhuvana ,  BIADSY, Fadi

IPC: G10L15/06 ,  G10L21/003 ,  G10L15/16

公开(公告)号：EP4407605A3

公开(公告)日：2024-10-23

申请号：EP24182760.9

申请日：2021-10-20

Applicant: Google LLC

Inventor： CHEN, Zhehuai ,  RAMABHADRAN, Bhuvana ,  ROSENBERG, Andrew ,  ZHANG, Yu ,  MENGIBAR, Pedro, J., Moreno

IPC: G10L13/02 ,  G10L15/06 ,  G10L15/16 ,  G10L13/047 ,  G10L13/00 ,  G10L13/08 ,  G10L13/10 ,  G10L13/033

Abstract: Disclosed herein is a computer-implemented for training a speech recognition model, the operations comprising: obtaining a multilingual text-to-speech (TTS) model; generating, using the multilingual TTS model, a native synthesized speech representation for an input text sequence in a first language that is conditioned on speaker characteristics of a native speaker of the first language; generating, using the multilingual TTS model, a cross-lingual synthesized speech representation for the input text sequence in the first language that is conditioned on speaker characteristics of a native speaker of a different second language; generating, using the speech recognition model, a first speech recognition result for the native synthesized speech representation and a second speech recognition result for the cross-lingual synthesized speech representation; determining a consistent loss term based on the first speech recognition result and the second speech recognition result; and updating parameters of the speech recognition model based on the consistent loss term.

公开(公告)号：EP4407605A2

公开(公告)日：2024-07-31

申请号：EP24182760.9

申请日：2021-10-20

Applicant: Google LLC

Inventor： CHEN, Zhehuai ,  RAMABHADRAN, Bhuvana ,  ROSENBERG, Andrew ,  ZHANG, Yu ,  MENGIBAR, Pedro, J., Moreno

IPC: G10L15/16

CPC classification number: G10L15/16 ,  G10L15/063 ,  G10L13/02

Abstract: Disclosed herein is a computer-implemented for training a speech recognition model, the operations comprising: obtaining a multilingual text-to-speech (TTS) model; generating, using the multilingual TTS model, a native synthesized speech representation for an input text sequence in a first language that is conditioned on speaker characteristics of a native speaker of the first language; generating, using the multilingual TTS model, a cross-lingual synthesized speech representation for the input text sequence in the first language that is conditioned on speaker characteristics of a native speaker of a different second language; generating, using the speech recognition model, a first speech recognition result for the native synthesized speech representation and a second speech recognition result for the cross-lingual synthesized speech representation; determining a consistent loss term based on the first speech recognition result and the second speech recognition result; and updating parameters of the speech recognition model based on the consistent loss term.

公开(公告)号：EP4088276A1

公开(公告)日：2022-11-16

申请号：EP21709828.4

申请日：2021-02-08

Applicant: Google LLC

Inventor： CHEN, Zhehuai ,  ROSENBERG, Andrew ,  RAMABHADRAN, Bhuvana ,  MORENO MENGIBAR, Pedro, J.

IPC: G10L15/06 ,  G10L13/00 ,  G10L15/16

公开(公告)号：EP4488993A2

公开(公告)日：2025-01-08

申请号：EP24215267.6

申请日：2022-03-22

Applicant: GOOGLE LLC

Inventor： GAUR, Neeraj ,  CHEN, Tongzhou ,  VARIANI, Ehsan ,  RAMABHADRAN, Bhuvana ,  HAGHANI, Parisa ,  MORENO MENGIBAR, Pedro J.

IPC: G10L15/16

Abstract: A method (400) includes receiving a sequence of acoustic frames (110) extracted from audio data corresponding to an utterance (106). During a first pass (301), the method includes processing the sequence of acoustic frames to generate N candidate hypotheses (204) for the utterance. During a second pass (302), and for each candidate hypothesis, the method includes: generating a respective un-normalized likelihood score (325); generating a respective external language model score (315); generating a standalone score (205) that models prior statistics of the corresponding candidate hypothesis; and generating a respective overall score (355) for the candidate hypothesis based on the un-normalized likelihood score, the external language model score, and the standalone score. The method also includes selecting the candidate hypothesis having the highest respective overall score from among the N candidate hypotheses as a final transcription (120) of the utterance.

公开(公告)号：EP4476723A1

公开(公告)日：2024-12-18

申请号：EP23717318.2

申请日：2023-03-21

Applicant: Google LLC

Inventor： WANG, Weiran ,  CHEN, Tongzhou ,  SAINATH, Tara N ,  VARIANI, Ehsan ,  PRABHAVALKAR, Rohit ,  HUANG, Ronny ,  RAMABHADRAN, Bhuvana ,  NEERAJ, Gaur ,  MAVANDADI, Sepand ,  PEYSER, Charles Caleb ,  STROHMAN, Trevor ,  HE, Yangzhang ,  RYBACH, David

IPC: G10L15/32 ,  G10L15/16

公开(公告)号：EP4361897A3

公开(公告)日：2024-07-17

申请号：EP24162746.2

申请日：2021-01-19

Applicant: GOOGLE LLC

Inventor： DATTA, Arindrima ,  RAMABHADRAN, Bhuvana ,  EMOND, Jesse ,  ROAK, Brian

IPC: G10L15/06 ,  G06N3/044 ,  G06N3/084 ,  G10L15/16 ,  G06F40/129 ,  G06F40/53

CPC classification number: G10L15/063 ,  G10L15/16 ,  G06F40/53 ,  G06F40/129 ,  G06N3/084 ,  G06N3/044

Abstract: A method (600) includes obtaining a plurality of training data sets (202) each associated with a respective native language and includes a plurality of respective training data samples (204). For each respective training data sample of each training data set in the respective native language, the method includes transliterating the corresponding transcription in the respective native script into corresponding transliterated text (121) representing the respective native language of the corresponding audio in a target script and associating the corresponding transliterated text in the target script with the corresponding audio (210) in the respective native language to generate a respective normalized training data sample (240). The method also includes training, using the normalized training data samples, a multilingual model (300) to predict speech recognition results (120) in the target script for corresponding speech utterances (106) spoken in any of the different native languages associated with the plurality of training data sets.