Sample labeling information auditing method and device

Abstract

The invention provides a sample labeling information auditing method and device. The method comprises the following steps: obtaining annotation samples needing to be audited and forming a training sample set; dividing the training sample set into a preset number of sub-sample sets, respectively training different sub-sample sets, and establishing different first recognition models; obtaining an identification sample set used for testing; carrying out identification on each identification sample in the identification sample set through the established different first identification models, andobtaining an identification result of each first identification model on the identification sample; counting the occurrence frequency of each identification result, and determining the first identification model corresponding to the identification result of which the occurrence frequency is smaller than a preset threshold as a target identification model when the identification result of which theoccurrence frequency is not smaller than the preset threshold exists; and carrying out labeling information audit on the labeled samples in the sub-sample set corresponding to the target identification model. By applying the scheme provided by the invention, the labeling result of the sample can be quickly audited.

Description

technical field [0001] The present invention relates to the technical field of artificial intelligence, in particular to a method, device, electronic equipment and computer-readable storage medium for reviewing sample labeling information. Background technique [0002] In the model training in the field of artificial intelligence, it is necessary to label samples, such as manually labeling samples, or automatically identifying and labeling samples through a pre-established neural network recognition model. In order to ensure the accuracy of model training, it is also necessary to review whether the annotation information of the samples is accurate. [0003] At present, the labeling information of all labeled samples is usually reviewed manually. However, due to the relatively large number of samples in the sample set, it will take more time and manpower to review the annotation information of the samples. Contents of the invention [0004] The object of the present inven...

AI Technical Summary

Problems solved by technology

However, due to the relatively large number of samples in the sample set, it ...

Images