US 11,928,439 B2
Translation method, target information determining method, related apparatus, and storage medium
Zhaopeng Tu, Shenzhen (CN); Hao Zhou, Shenzhen (CN); and Shuming Shi, Shenzhen (CN)
Assigned to TENCENT TECHNOLOGY (SHENZHEN) COMPANY LIMITED, Shenzhen (CN)
Filed by TENCENT TECHNOLOGY (SHENZHEN) COMPANY LIMITED, Shenzhen (CN)
Filed on Jan. 22, 2020, as Appl. No. 16/749,243.
Application 16/749,243 is a continuation of application No. PCT/CN2018/095231, filed on Jul. 11, 2018.
Claims priority of application No. 201710612833.7 (CN), filed on Jul. 25, 2017.
Prior Publication US 2020/0226328 A1, Jul. 16, 2020
Int. Cl. G06F 40/58 (2020.01); G06N 3/04 (2023.01); G06N 3/08 (2023.01)
CPC G06F 40/58 (2020.01) [G06N 3/04 (2013.01)] 20 Claims
OG exemplary drawing
 
1. A target information determining method, comprising:
encoding to-be-processed text information, to obtain a source vector representation sequence;
obtaining a source context vector corresponding to a first instance according to the source vector representation sequence, wherein the source context vector corresponding to the first instance indicates to-be-processed source content in the to-be-processed text information at the first instance;
obtaining a past translation vector corresponding to a second instance that represents a translated source content in the source vector representation sequence at the second instance, a future translation vector corresponding to the second instance that represents an untranslated source content in the source vector representation sequence at the second instance, and the second instance is an adjacent instance before the first instance;
determining a past translation vector corresponding to the first instance and a future translation vector corresponding to the first instance according to the source context vector corresponding to the first instance, the past translation vector corresponding to the second instance, and the future translation vector corresponding to the second instance; and
inputting the past translation vector corresponding to the second instance, the future translation vector corresponding to the first instance and the source context vector corresponding to the first instance to a decoder, to obtain target information of the first instance, wherein the decoder includes two hidden layers respectively directed to the past translation vector corresponding to the second instance and the future translation vector corresponding to the first instance.