DATA SAMPLE LABEL PROCESSING METHOD AND APPARATUS

Title:

DATA SAMPLE LABEL PROCESSING METHOD AND APPARATUS

Document Type and Number:

WIPO Patent Application WO/2019/095782

Kind Code:

A1

Abstract:

Disclosed are a data sample label processing method and apparatus. The data sample label processing method comprises: obtaining an indeterminate label sample set and a determined label sample set; performing iteration processing with the following steps, until an accuracy rate of a learning label satisfies a preset requirement: combining the current indeterminate label sample set and the determined label sample set, and training same to obtain a prediction model; inputting the indeterminate label sample into the prediction model, and determining a prediction value output by said model as the learning label of the indeterminate label sample; obtaining a sampling subset having the learning label, and checking the learning label of the sampling subset, so as to obtain the accuracy rate of the learning label, and if the accuracy rate of the learning label does not satisfy a preset requirement, determining a label checking result as a labeled result of the sampling subset, and moving the sampling subset from the indeterminate label sample set to the determined label sample set; after iteration ends, determining the current learning label of the indeterminate label sample as the labeled result.

Inventors:

CHEN FAN (CN)
QI XIANG (CN)
WANG DESHENG (CN)
WANG HANBIN (CN)
GUO QILIN (CN)

Application Number:

PCT/CN2018/102732

Publication Date:

May 23, 2019

Filing Date:

August 28, 2018

Export Citation:

Click for automatic bibliography generation Help

Assignee:

ALIBABA GROUP HOLDING LTD (CN)
CHEN FAN (CN)
QI XIANG (CN)
WANG DESHENG (CN)
WANG HANBIN (CN)
GUO QILIN (CN)

International Classes:

G06K9/62

Foreign References:

CN108021931A	2018-05-11
CN104166706A	2014-11-26
CN107016406A	2017-08-04
US20160034788A1	2016-02-04

Other References: