Transducer-based streaming push for cascaded encoders

A method (400) includes receiving a sequence of acoustic frames (110) and generating, by a first encoder (210), a first higher order feature representation (212) for a corresponding acoustic frame in the sequence of acoustic frames. The method further includes generating, by the first pass transducer decoder (201), a first pass speech recognition hypothesis (120a) for the corresponding first high-order feature representation, and generating, by the text encoder (240), a text code (242) for the corresponding first pass speech recognition hypothesis. The method also includes generating, by a second encoder (220), a second high-order feature representation (222) for the corresponding first high-order feature representation. The method further includes generating, by a second pass transducer decoder (202), a second pass speech recognition hypothesis (120b) using the corresponding second higher order feature representation and the corresponding text coding. 一种方法(400)包括接收声学帧(110)的序列，并且由第一编码器(210)针对声学帧序列中的对应的声学帧生成第一