Label prediction method and device, electronic equipment and medium

Abstract

The invention discloses a label prediction method and device, electronic equipment and a medium. The label prediction method comprises the steps: training a similarity evaluation model for label prediction; inputting the to-be-predicted data into the similarity evaluation model for similarity evaluation; performing label prediction based on the evaluated similarity to obtain a prediction label corresponding to the to-be-predicted data; training a similarity evaluation model for label prediction, including: constructing a query set and a support set; inputting a query sample and each support sample in each support subset into the similarity evaluation model to obtain the similarity between each support sample and the query sample; obtaining a label prediction set of the support set for thequery sample according to the similarity between each support sample and the query sample and the label corresponding to each support sample; and correcting the similarity evaluation model according to the difference between the label prediction set and the label carried by the query sample. Based on the corrected model, more accurate similarity can be obtained, and the accuracy of label prediction can be improved.

Description

technical field [0001] The present invention relates to the technical field of Internet communication, in particular to a label prediction method, device, electronic equipment and medium. Background technique [0002] The corresponding tags can be marked on the data to be predicted, and the tags can reflect a certain degree of portrait description of the data to be predicted, and the use of tags can provide convenience for the corresponding scenes where the data to be predicted exists. For example, labeling the data to be predicted pointing to a picture with a corresponding label can facilitate searching for other pictures belonging to the same category or similar to the picture. Labeling the data to be predicted pointing to the user account with a corresponding label can facilitate pushing information of a corresponding category to the user account. [0003] In the prior art, the label prediction of the data to be predicted is often realized based on the relationship betwe...

AI Technical Summary

Problems solved by technology

[0004] In order to solve the problem of low label accuracy in label prediction applied in the prior a

Images