Looking for breakthrough ideas for innovation challenges? Try Patsnap Eureka!

Method for extracting text sentence features

A sentence and text technology, applied in the field of extracting text sentence features, can solve the problem of low accuracy of deduplication

Pending Publication Date: 2020-10-16
CHINA JILIANG UNIV
View PDF0 Cites 0 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0004] A method for extracting features of text sentences provided by the present invention aims to solve the problem of low accuracy in deduplication existing in the prior art

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
  • Method for extracting text sentence features
  • Method for extracting text sentence features
  • Method for extracting text sentence features

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0057] Such as figure 1 As shown, a method for extracting text sentence features includes the following steps:

[0058] S110. Calculate the word vector according to the Self-Attention algorithm to obtain an attention sequence;

[0059] S120. Perform a convolution operation on the attention sequence to obtain a feature matrix;

[0060] S130. Output text sentence features using a maximum pooling algorithm on the feature matrix.

[0061] According to Embodiment 1, it can be seen that firstly, the balance operation is performed on the unbalanced samples. The balance operation technology in this method is to effectively combine undersampling and oversampling to obtain the word vector of the balanced sample, and perform Self-Attention algorithm calculation on the word vector. Perform convolution operation on the obtained attention sequence to obtain the feature matrix, and finally calculate the feature matrix according to the maximum pooling algorithm to output text sentence featu...

Embodiment 2

[0063] Such as figure 2 As shown, a method of extracting text sentence features, including:

[0064] S210. Calculate the word vector according to the Self-Attention algorithm to obtain an attention sequence;

[0065]S220. Obtain the weight of the word vector by using a dot product algorithm;

[0066] S230. Perform normalization processing on the weights;

[0067] S240. Output the attention sequence according to the weighted sum of the weight Key and the key value Value, and the calculation formula is:

[0068] where Q∈R n×dk ,K∈R m×dk , V ∈ R m×dv .

[0069] According to Example 2, the similarity between Query and each Key is calculated according to the dot product similarity function, and the weight of the word vector is obtained, and then the weight is normalized according to the Softmax function, and finally the weight and the corresponding key Value Weighted summation to get the attention sequence, the calculation formula is where Q∈R n×dk ,K∈R m×dk , V ∈ R ...

Embodiment 3

[0071] Such as image 3 As shown, a method of extracting text sentence features, including:

[0072] S310. Calculate the word vector according to the Self-Attention algorithm to obtain an attention sequence;

[0073] S320. Perform a convolution operation on the attention sequence to obtain a feature matrix;

[0074] S330. Process the attention sequence into an attention component according to the length of the convolution kernel;

[0075] S340. Perform a convolution operation on the attention component to output a feature matrix, and the calculation formula is: C=(c 1 ,c 2 ,...,c n-h+1 ), where c i is the attention component X i :i+h-1 features extracted after convolution operation.

[0076] In Embodiment 3, the attention sequence is processed as an attention component according to the length of the convolution kernel. For example, a convolution kernel with a length of h can divide the attention sequence into {X0:h-1, X1:h,..., Xi:i+h-1,...,Xn-h+1:n} style attention co...

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 method for extracting text sentence features, which comprises the following steps of: performing Self-Attention algorithm calculation on a word vector, performing convolutionoperation on an obtained attention sequence to obtain a feature matrix, and finally performing calculation on the feature matrix according to a maximum pooling algorithm to output text sentence features. According to the method, the problems of low duplicate removal accuracy and disordered word sorting in the prior art are solved.

Description

technical field [0001] The invention relates to the field of sentiment analysis, in particular to a method for extracting features of text sentences. Background technique [0002] With WeChat, Alipay, Weibo and other social platforms and shopping platforms having a huge impact on people's lives, more and more users communicate and share their lives on the platforms through language and text. Semantic analysis of these language characters can obtain the user's inner activities and analyze the user's personality characteristics. [0003] The premise of semantic analysis is to extract text sentence features. The existing methods for extracting text sentence features have the problems of low word deduplication accuracy and chaotic order of words. Contents of the invention [0004] A method for extracting features of text sentences provided by the present invention aims to solve the problem of low accuracy in deduplication in the prior art. [0005] To achieve the above objec...

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): G06F40/30G06F40/216
CPCG06F40/30G06F40/216
Inventor 杨小兵陈欣
Owner CHINA JILIANG UNIV
Who we serve
  • R&D Engineer
  • R&D Manager
  • IP Professional
Why Patsnap Eureka
  • Industry Leading Data Capabilities
  • Powerful AI technology
  • Patent DNA Extraction
Social media
Patsnap Eureka Blog
Learn More
PatSnap group products

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

© 2024 PatSnap. All rights reserved.Legal|Privacy policy|Modern Slavery Act Transparency Statement|Sitemap|About US| Contact US: help@patsnap.com