Data set tagging method and related apparatus

Abstract

The invention discloses a data set tagging method. The method comprises the steps of selecting untagged data from original data according to a preset rule to obtain a candidate data set; performing uncertainty analysis on the candidate data set, and performing to-be-tagged data screening according to an analysis result, thereby obtaining a to-be-tagged data set; and performing tagging processing on the to-be-tagged data set according to received tagging information, thereby obtaining a tagged data set. The data set is subjected to data screening according to the uncertainty to obtain an uncertain data set suitable for model processing; the data set is tagged, so that the model training and testing efficiency can be improved, a better effect can be achieved by less data, and the overall efficiency of supervised learning is improved. The invention furthermore discloses a data set tagging apparatus, a server and a computer readable storage medium, which have the abovementioned beneficialeffects.

Description

technical field [0001] The present application relates to the field of computer technology, and in particular to a data set labeling method, labeling device, server and computer-readable storage medium. Background technique [0002] With the development of information technology, machine learning technology has been applied to more and more fields to improve the efficiency of dealing with problems in different application scenarios. Machine learning is mainly to train through a large amount of data to obtain a more accurate recognition model. At the same time, it is necessary to continuously use the original data to test the recognition model to judge whether the recognition model meets the learning requirements. [0003] The current mainstream machine learning is still supervised learning, and labeled data is indispensable in supervised learning. With the further development of the Internet, a large amount of data is generated every day, and these data are messy, unlabeled...

AI Technical Summary

Problems solved by technology

[0005] However, in the general manual labeling process, it is inevitable to label a large number of data that can be recognized by existing recognition models. These data that can be recognized by the recognition model cannot make the recognition model better tested, that is, it is impossible to achieve the recognition model. The expected effect of labeling test data reduces the overall training efficiency of supervised learning

Images