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

Sentence noise design method, equipment and computer storage medium

A design method and noise technology, applied in computing, neural learning methods, instruments, etc., can solve the problem of low fluency of noisy text

Active Publication Date: 2021-04-09
PENG CHENG LAB
View PDF12 Cites 0 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0009] In view of this, a sentence noise design method, equipment and computer storage medium are provided to solve the problem of low fluency of noise text

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
  • Sentence noise design method, equipment and computer storage medium
  • Sentence noise design method, equipment and computer storage medium
  • Sentence noise design method, equipment and computer storage medium

Examples

Experimental program
Comparison scheme
Effect test

no. 1 example

[0071] refer to figure 2 , figure 2 For the first embodiment of the sentence noise design method of the present invention, the method includes:

[0072] Step S110: Preprocessing the original text to generate the first noise text.

[0073] The original text may be a text in a preloaded corpus, or a text in any corpus, which is not limited here.

[0074] Preprocessing may be a preparation process performed before the original text is generated into the first noise text.

[0075] The first noise text may be a text formed by adding noise words to the original text, and the first noise text also provides position information of the noise words in the first noise text for subsequent fluency processing.

[0076] Step S120: Calculate the sentence structure similarity between the first noise text and the text in the preloaded corpus based on the adaptive sliding window, and perform fluency optimization processing on the first noise text by using the sentence structure similarity, ...

no. 2 example

[0155] refer to Figure 11 , Figure 11 For the second embodiment of the sentence noise design method of the present invention, the method also includes:

[0156] Step S210: Preprocessing the original text to generate the first noise text.

[0157] Step S220: Execute fluency optimization processing on the first noise text to obtain a second noise text whose fluency meets a preset condition.

[0158] Step S230: Use the deep learning model to predict the second noise text, and if the predicted value is different from the prediction value of the original text using the deep learning model, take the second noise text as the target result.

[0159] Step S240: If the predicted value is the same as the predicted value of the original text using the deep learning model, re-execute the generating process of the first noise text.

[0160] The predicted value is the same as the predicted value of the original text using the deep learning model, indicating that the fluency optimization...

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 sentence noise design method, equipment and computer storage medium. The method includes the following steps: preprocessing the original text to generate a first noise text; Load the sentence structure similarity between the texts in the corpus, and use the sentence structure similarity to perform fluency optimization processing on the first noise text, and obtain the second noise text whose fluency meets the preset conditions; using a deep learning model Predict the second noisy text, and if the predicted value is different from the predicted value of the original text using the deep learning model, then use the second noisy text as the target result. The present invention solves the problem of low fluency of noise texts, implements an iterative-based positioning and noise word injection attack, and adds a generated noise fluency optimization algorithm, so that generated noise texts are more in line with correct grammar and human reading habits.

Description

technical field [0001] The invention relates to the field of natural language processing, in particular to a sentence noise design method, equipment and computer storage medium. Background technique [0002] Adversarial examples refer to deliberately adding subtle disturbances to the input samples, causing the model to output a wrong result with high confidence. It has achieved some results in the field of image and speech, but in the field of text, due to its discrete nature, it still faces many challenges. . For the natural language processing attack model, it is not only necessary to be able to deceive the target model, but also the adversarial samples generated by it need to meet three attributes [0003] (1) Consistency of human predictions, that is, human predictions of the input text remain unchanged; [0004] (2) Semantic similarity, the generated adversarial examples should maintain similar meanings to the original text as much as possible. [0005] (3) Sentence ...

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 Patents(China)
IPC IPC(8): G06F40/247G06F40/211G06F40/194G06F40/289G06N3/04G06N3/08
CPCG06F40/247G06F40/194G06N3/08G06F40/211G06F40/289G06N3/044G06N3/045
Inventor 杨孙傲钟晓雄张伟哲周颖程正涛
Owner PENG CHENG LAB