Check patentability & draft patents in minutes with Patsnap Eureka AI!

Genetic algorithm-based anti-convolutional neural network sentence similarity calculation method

A convolutional neural network and sentence similarity technology, applied in the field of algorithm programs, can solve problems such as model recognition and judgment ability cannot be guaranteed, lack of deep learning model interaction, etc.

Pending Publication Date: 2021-06-04
CHINA UNIV OF PETROLEUM (EAST CHINA)
View PDF0 Cites 1 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0006] The purpose of the present invention is to solve the problem that the training of adversarial samples lacks the interaction with the deep learning model, and the ability of the model to identify and judge samples cannot be guaranteed

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
  • Genetic algorithm-based anti-convolutional neural network sentence similarity calculation method
  • Genetic algorithm-based anti-convolutional neural network sentence similarity calculation method
  • Genetic algorithm-based anti-convolutional neural network sentence similarity calculation method

Examples

Experimental program
Comparison scheme
Effect test

Embodiment Construction

[0029] The accompanying drawings are for illustrative purposes only and should not be construed as limiting the patent.

[0030] The present invention will be further elaborated below in conjunction with the accompanying drawings and implementation examples.

[0031] figure 1 Flowchart for adversarial example generation. The process of generating an adversarial sample using a genetic algorithm is divided into three steps. ① After the sample is word-segmented, the input sample of the genetic algorithm is obtained, and the sample is a word in the sentence. ② Samples are iterated according to the grammatical rules and part-of-speech rules in accordance with the crossover, genetic and mutation processes of the genetic algorithm. ③The samples after iteration are input into the convolutional neural network model to judge whether they meet the pre-set threshold conditions. If it is satisfied, the iterative process ends to obtain the input of the adversarial example. If not, proc...

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 an adversarial convolutional neural network sentence similarity calculation method based on a genetic algorithm, and belongs to the field of Chinese natural language processing. The method aims at solving the problem that an existing method lacks an adversarial sample detection mechanism and cannot guarantee model safety. Based on the genetic algorithm and the adversarial convolutional neural network, the safety of the sentence similarity calculation model is improved by setting the modification rate of the text and interacting with the deep learning model. Parameters such as synonyms, position information and change rate of texts are considered when adversarial samples are generated in the model. According to the method, a large number of experiments are performed on a Microsoft paraphrasing corpus to verify the effectiveness of the method, and the method can be used for improving the safety of a sentence similarity calculation model.

Description

technical field [0001] The invention belongs to an algorithm program and relates to a security mechanism of a text processing model, mainly the generation and detection of adversarial samples of a sentence semantic evaluation model. Background technique [0002] After data training, the text processing model can recognize the semantics of the text and then analyze a large amount of text data. The deep learning model completes the semantic classification task of the text according to the characteristics of the input text through the pre-trained parameters. However, the result of interfering with the output of the model by converting the sentence into synonyms or word order poses a challenge to the security of the deep learning model. This type of text that has undergone synonym replacement or word order transformation is called an adversarial example. In the traditional security enhancement model, the text processing model is trained by adding adversarial samples to improve ...

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
IPC IPC(8): G06K9/62G06F16/35G06N3/04G06N3/08
CPCG06F16/35G06N3/086G06N3/045G06F18/22
Inventor 黄兴哲高亚
Owner CHINA UNIV OF PETROLEUM (EAST CHINA)
Features
  • R&D
  • Intellectual Property
  • Life Sciences
  • Materials
  • Tech Scout
Why Patsnap Eureka
  • Unparalleled Data Quality
  • Higher Quality Content
  • 60% Fewer Hallucinations
Social media
Patsnap Eureka Blog
Learn More

Browse by: Latest US Patents, China's latest patents, Technical Efficacy Thesaurus, Application Domain, Technology Topic, Popular Technical Reports.

© 2025 PatSnap. All rights reserved.Legal|Privacy policy|Modern Slavery Act Transparency Statement|Sitemap|About US| Contact US: help@patsnap.com