View source for 18815537. TWO-PASS END TO END SPEECH RECOGNITION (GOOGLE LLC)

=TWO-PASS END TO END SPEECH RECOGNITION=


==Organization Name==

[[:Category:GOOGLE LLC|GOOGLE LLC]]


[[Category:GOOGLE LLC]]

==Inventor(s)==

[[:Category:Tara N. Sainath of Jersey City NJ (US)|Tara N. Sainath of Jersey City NJ (US)]][[Category:Tara N. Sainath of Jersey City NJ (US)]]

[[:Category:Yanzhang He of Palo Alto CA (US)|Yanzhang He of Palo Alto CA (US)]][[Category:Yanzhang He of Palo Alto CA (US)]]

[[:Category:Bo Li of Fremont CA (US)|Bo Li of Fremont CA (US)]][[Category:Bo Li of Fremont CA (US)]]

[[:Category:Arun Narayanan of Milpitas CA (US)|Arun Narayanan of Milpitas CA (US)]][[Category:Arun Narayanan of Milpitas CA (US)]]

[[:Category:Ruoming Pang of New York NY (US)|Ruoming Pang of New York NY (US)]][[Category:Ruoming Pang of New York NY (US)]]

[[:Category:Antoine Jean Bruguier of Milpitas CA (US)|Antoine Jean Bruguier of Milpitas CA (US)]][[Category:Antoine Jean Bruguier of Milpitas CA (US)]]

[[:Category:Shuo-yiin Chang of Mountain View CA (US)|Shuo-yiin Chang of Mountain View CA (US)]][[Category:Shuo-yiin Chang of Mountain View CA (US)]]

[[:Category:Wei Li of Fremont CA (US)|Wei Li of Fremont CA (US)]][[Category:Wei Li of Fremont CA (US)]]

==TWO-PASS END TO END SPEECH RECOGNITION==


This abstract first appeared for US patent application 18815537 titled 'TWO-PASS END TO END SPEECH RECOGNITION




==Original Abstract Submitted==

Two-pass automatic speech recognition (ASR) models can be used to perform streaming on-device ASR to generate a text representation of an utterance captured in audio data. Various implementations include a first-pass portion of the ASR model used to generate streaming candidate recognition(s) of an utterance captured in audio data. For example, the first-pass portion can include a recurrent neural network transformer (RNN-T) decoder. Various implementations include a second-pass portion of the ASR model used to revise the streaming candidate recognition(s) of the utterance and generate a text representation of the utterance. For example, the second-pass portion can include a listen attend spell (LAS) decoder. Various implementations include a shared encoder shared between the RNN-T decoder and the LAS decoder.



[[Category:G10L15/16]]
[[Category:G06N3/08]]
[[Category:G10L15/05]]
[[Category:G10L15/06]]
[[Category:G10L15/22]]
[[Category:CPC_G10L15/16]]