公开(公告)号：EP4062400B1

公开(公告)日：2024-07-17

申请号：EP21706060.7

申请日：2021-01-14

IPC分类号： G10L15/16 ,  G10L15/32 ,  G06N3/04 ,  G10L15/02 ,  G06N3/044 ,  G06N3/088 ,  G10L15/08 ,  G06N3/084

CPC分类号： G10L15/16 ,  G10L2015/08120130101 ,  G10L2015/08520130101 ,  G10L15/02 ,  G10L15/32 ,  G06N3/088 ,  G06N3/084 ,  G06N3/044

公开(公告)号：EP4414896A2

公开(公告)日：2024-08-14

申请号：EP24184116.2

申请日：2021-01-14

申请人： GOOGLE LLC

发明人： HU, Ke ,  SAINATH, Tara, N ,  PANG, Ruoming ,  PRABHAVALKAR, Rohit, Prakash

IPC分类号： G06N3/088

CPC分类号： G10L15/16 ,  G10L2015/08120130101 ,  G10L2015/08520130101 ,  G10L15/02 ,  G10L15/32 ,  G06N3/088 ,  G06N3/084 ,  G06N3/044

摘要： Disclosed herein is a computer-implemented method when executed on data processing hardware causes the data processing hardware to perform operations comprising:
receiving a sequence of audio features characterizing an utterance;
based on the sequence of audio features, generating, using a first-pass decoder model, a plurality of first-pass speech recognition hypotheses, each first-pass speech recognition hypothesis corresponding to a candidate transcription of the utterance;
generating, using a long short-term memory (LSTM) encoder, a first-pass encoding of the plurality of first-pass speech recognition hypotheses; and
based on the sequence of audio features and the first-pass encoding, generating, using a second-pass decoder model, a second-pass hypothesis that rescores the plurality of first-pass speech recognition hypotheses.