Data labeling method, device and system and storage medium

Abstract

The embodiment of the invention provides a data labeling method, device and system and a storage medium. The data labeling method includes: obtaining first quantity of unlabeled data and pre-labelinginformation thereof, wherein the pre-labeling information is obtained by using a labeling model to pre-label the first quantity of unlabeled data, and includes pre-labeling results; displaying the first quantity of unlabeled data and the pre-labeling results thereof on a display interface; receiving first feedback information of a user on the first quantity of unlabeled data; and determining finallabeling results of the first quantity of unlabeled data according to the first feedback information. According to the data labeling method, device and system and the storage medium of the embodimentof the invention, firstly, the unlabeled data are pre-labeled by the data labeling system, the labeled data and the pre-labeling results thereof can be displayed on the display interface, the user only needs to change wrong pre-labeling results, thus labeling efficiency can be greatly improved, and labeling costs can be reduced.

Description

technical field [0001] The present invention relates to the field of computer technology, and more particularly to a data labeling method, device, system and storage medium. Background technique [0002] With the development of artificial intelligence today, the role of data is becoming more and more prominent. Training a neural network model usually requires millions or even hundreds of millions of data. The labeling cycle and cost of data directly affect the industry competitiveness of an artificial intelligence company. [0003] The data labeling process of the current data labeling platform is manual labeling one by one, and the labeling interface based on this labeling process is also labeling single data points one by one. The current data labeling platform has the following deficiencies: the data labeling mode is to manually label the data one by one; the labeling cost is generally proportional to the size of the data set, and when labeling a very large data set, it...

AI Technical Summary

Problems solved by technology

The current data labeling platform has the following deficiencies: the data labeling mode is to manually label the data one by one; the labeling cost is generally proportional to the size of the data set, and it usually requires a large human input and a long labeling time when labeling a very large data set cycle

Images