User label generation method and device and computer readable storage medium

Abstract

The embodiment of the invention discloses a user label generation method and device and a computer readable storage medium. The method comprises the steps that a behavior text sequence of a user is acquired, the behavior text sequence is randomly initialized by using a bidirectional long short-term memory network BiLSTM to obtain a first feature vector, and the behavior text sequence is a sequencecomposed of texts which are generated according to a time sequence and used for representing user behaviors; social network information of the user is acquired, and the social network information isprocessed according to an inductive learning algorithm to obtain a second feature vector; social statistics information of the user is acquired and a third feature vector is acquired according to thesocial statistics information; and the first feature vector, the second feature vector and the third feature vector are transversely spliced to obtain the user label of the user. The embodiment of theinvention is beneficial to improving the precision of generating the user label.

Description

technical field [0001] The present application relates to the technical field of artificial intelligence, and in particular to a user label generation method, device and computer-readable storage medium. Background technique [0002] At present, multimodal information fusion technology is a research hotspot in the field of artificial intelligence. For example, in the process of user label production, it is often necessary to organically integrate different modal data such as images, texts, social networks, geographic locations, and social statistical information. Together, we can construct user behavior characteristics from multiple angles and in all directions, give accurate user labels, and then give accurate prediction values, so as to better serve product operations, personalized recommendations, and precise advertising. [0003] In addition, social network information plays a pivotal role in user tags, especially for users with sparse features. Other information about ...

AI Technical Summary

Problems solved by technology

[0004] At present, the linear splicing method can be used when fusing social network information with other information of other modalities. However, for ultra-large-scale social network scenarios, the social relationships of each user are intricate, and it is impossible to generate each user's social network. Network vector, in addition, because different modal information contributes differently to the user tags to which the user belongs, and only simple splicing is performed, the generated user tags have low accuracy

Images