-
公开(公告)号:EP3620940A1
公开(公告)日:2020-03-11
申请号:EP18814267.3
申请日:2018-06-06
发明人: CHANG, Xiaofu , CHAO, Linlin , XU, Peng , LI, Xiaolong
IPC分类号: G06F17/30
摘要: A dialog generation method, apparatus, and electronic device are disclosed in this application. The method includes training a sequence to sequence (seq2seq)-based dialog model using a loss function including topic range constraint information; and generating a dialog using the trained dialog model. With embodiments of this application, topic range constraint information is introduced in the process of dialog model training using a loss function including the topic range constraint information, thus helping to prevent the trained model from producing low-quality meaningless replies.
-
2.
公开(公告)号:EP3690787A1
公开(公告)日:2020-08-05
申请号:EP18888429.0
申请日:2018-10-18
发明人: SONG, Le , LI, Hui , GE, Zhibang , HUANG, Xin , WEN, Chunyang , WANG, Lin , JIANG, Tao , WANG, Yiguang , CHANG, Xiaofu , ZHU, Guanyin
摘要: Implementations of the present specification disclose a graphical structure model-based method, apparatus, and device for credit risk control. The solution includes: establishing an appropriate enterprise relationship network in advance based on data related to an enterprise and factors associated with the enterprise; performing feature integration and defining a graphical structure model; training the graphical structure model by using a labeled sample; calculating embedding vectors of a node after a plurality of iterations in hidden feature space and a predicted probability of the node by using the trained graphical structure model; and performing credit risk control on the node based on the embedded factor and the predicted probability.
-