Unlock instant, AI-driven research and patent intelligence for your innovation.

Label correction method and system

A label and original label technology, applied in neural learning methods, biological neural network models, semantic analysis, etc., can solve problems such as reducing manual participation, and achieve the effects of reducing manual participation, optimizing text imbalance, and quickly repairing wrong labels.

Pending Publication Date: 2021-09-24
EMOTIBOT TECH LTD
View PDF3 Cites 0 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0003] Aiming at the deficiencies in the prior art, the present invention provides a method and system for label correction, which uses a semantic classification model and an interactive verification method to analyze text label labels and correct wrong labels. The correction process reduces manual participation and can Quickly fix wrong labels and optimize text imbalance

Method used

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
View more

Image

Smart Image Click on the blue labels to locate them in the text.
Viewing Examples
Smart Image
  • Label correction method and system
  • Label correction method and system
  • Label correction method and system

Examples

Experimental program
Comparison scheme
Effect test

example 1

[0073] A label correction method includes the following steps:

[0074] Step S1:

[0075] Get all texts: ["Shanghai weather", "Shenzhen weather", "Beijing weather", "play a song", "want to listen to a song"];

[0076] All text original tags: ["check the weather", "check the weather", "listen to music", "listen to music", "listen to music"].

[0077] Step S2:

[0078] Set k=2, perform 2-fold interactive verification.

[0079] fold 1:

[0080] Training text: ["Shenzhen weather", "Beijing weather", "want to listen to songs"];

[0081] The original label of the training text: ["check the weather", "listen to music", "listen to music"];

[0082] Verification text: ["Shanghai weather", "play a song"], verification text original label: ["check the weather", "listen to music"].

[0083] Train the model and predict the validation text

[0084] Verify the text prediction label: ["check the weather", "listen to music"];

[0085] The confidence score of the verification text predi...

example 2

[0107] step 1:

[0108] All text: ["fun", "happy", "bad attitude", "unhappy"];

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More

PUM

No PUM Login to View More

Abstract

The invention discloses a label correction method and system, and the method comprises the following steps: S1, obtaining all texts and an original label of each text, each label having at least two corresponding texts; s2, predicting all texts by adopting an interactive verification method to obtain a prediction result; s3, calculating the accuracy according to a prediction result; s4, processing the prediction result of each text; and S5, repeatedly executing the steps S2 to S4 until the calculated accuracy value is unchanged, and obtaining the corrected label and the corresponding text. According to the method, a semantic classification model and an interactive verification method are adopted to analyze annotation labels of a text, if the labels have errors, correction is carried out, manual participation is reduced in the correction process, the error labels can be rapidly repaired, and meanwhile the problem of text imbalance is solved.

Description

technical field [0001] The present invention relates to the technical field of data processing methods, and in particular to a label correction method and system. Background technique [0002] Wrongly labeled text is often encountered when training semantic classification models, which affects model accuracy. However, manually reviewing and relabeling texts is labor-intensive and time-consuming. Therefore, there is an urgent need for methods to automatically repair labeling errors. Contents of the invention [0003] Aiming at the deficiencies in the prior art, the present invention provides a method and system for label correction, which uses a semantic classification model and an interactive verification method to analyze text label labels and correct wrong labels. The correction process reduces manual participation and can Quickly fix wrong labels and optimize text imbalance. [0004] In the first aspect, a label correction method provided by an embodiment of the pres...

Claims

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More

Application Information

Patent Timeline
no application Login to View More
Patent Type & Authority Applications(China)
IPC IPC(8): G06F16/35G06F40/30G06K9/62G06N3/04G06N3/08
CPCG06F16/35G06F40/30G06N3/08G06N3/045G06F18/214
Inventor 简仁贤佘昌宪
Owner EMOTIBOT TECH LTD