Methods, apparatuses, devices, and media for text processing

Abstract

The present disclosure provides a text processing method, device, equipment and medium, relating to the field of artificial intelligence. The method comprises: processing embedding features of a plurality of input words by using a first self-attention layer to obtain a plurality of first vectors corresponding to a source text and at least one second vector corresponding to a target text; calculating a first correlation between a plurality of first words and a plurality of first expert sub-networks to determine a first word matching each first expert sub-network in the plurality of first words, and generating a third vector of the corresponding first word by using the first expert sub-network; calculating a second correlation between at least one second word and a plurality of second expert sub-networks to determine a second expert sub-network matching each second word in the plurality of second expert sub-networks, and generating a fourth vector of the corresponding second word by using the second expert sub-network; and generating an expanded target text based on the plurality of third vectors and the plurality of fourth vectors.

Images