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

A Cross-Domain Semantic Information Retrieval Method Based on Convolutional Neural Network

A convolutional neural network and semantic information technology, applied in the field of computer natural language processing, can solve problems such as high real-time computing power requirements, affecting model accuracy, unsatisfactory effect, etc., to overcome sparse features, improve classification accuracy, fast effect

Active Publication Date: 2021-06-18
KUNMING UNIV OF SCI & TECH
View PDF0 Cites 0 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

The word feature method for sentence similarity calculation generally relies on constructing a vector space, but the obvious defect of this type of method is that the features are sparse, and the effect is not ideal when used on a slightly larger corpus
In order to solve the problem of feature sparsity, Wang (, 2006) proposed a sentence similarity calculation based on lexical decomposition and combination, which vectorizes the compared sentences, decomposes the formed sentence feature matrix, and uses it to approximate Statement calculation, but this method requires high real-time computing capabilities of the environment
The method of word meaning features mainly relies on external semantic dictionaries, such as the HowNet-based information amount calculation semantic similarity algorithm proposed by You (, 2013), but this kind of method has too strong limitations, and the integrity of external semantic dictionaries directly affect the accuracy of the model
The method of calculating sentence similarity based on syntactic analysis features, such as Li (, 2013), proposed the method of calculating similarity of Chinese sentences based on frame semantic analysis, which mainly uses dependency relationship to extract core words and construct similarity matrix for similarity Degree calculation, this kind of method actually stays in the analysis of shallow word meaning, ignores the relationship between words in the sentence when calculating similarity, and the effect is not ideal in short text analysis

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
  • A Cross-Domain Semantic Information Retrieval Method Based on Convolutional Neural Network
  • A Cross-Domain Semantic Information Retrieval Method Based on Convolutional Neural Network
  • A Cross-Domain Semantic Information Retrieval Method Based on Convolutional Neural Network

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0036] Embodiment 1: as Figure 1-4 As shown, a convolutional neural network-based cross-domain semantic information retrieval method, the specific steps of the method are as follows:

[0037] Step1. First, preprocess the retrieved information in the knowledge base, and then represent the processed information as sentence vectors. Then, use the SVM classifier of LibSVM to classify and train the sentence vectors to obtain the SVM classification model, and use the classification model to classify the user Classify the retrieved information to obtain the category of the user retrieved information;

[0038] Step2. Convert the user retrieval information into a sentence feature matrix. According to the category determined in Step1, use the corresponding approximate semantic convolutional neural network retrieval model for processing. The approximate semantic convolutional neural network retrieval model conversion layer will generalize the user retrieval information. The features of...

Embodiment 2

[0044] Embodiment 2: as Figure 1-4 As shown, a convolutional neural network-based cross-domain semantic information retrieval method, the specific steps of the method are as follows:

[0045] Step1. First, preprocess the retrieved information in the knowledge base, and then represent the processed information as sentence vectors. Then, use the SVM classifier of LibSVM to classify and train the sentence vectors to obtain the SVM classification model, and use the classification model to classify the user Classify the retrieved information to obtain the category of the user retrieved information;

[0046] Step2. Convert the user retrieval information into a sentence feature matrix. According to the category determined in Step1, use the corresponding approximate semantic convolutional neural network retrieval model for processing. The approximate semantic convolutional neural network retrieval model conversion layer will generalize the user retrieval information. The features of...

Embodiment 3

[0066] Embodiment 3: as Figure 1-4As shown, a convolutional neural network-based cross-domain semantic information retrieval method, the specific steps of the method are as follows:

[0067] Step1. First, preprocess the retrieved information in the knowledge base, and then represent the processed information as sentence vectors. Then, use the SVM classifier of LibSVM to classify and train the sentence vectors to obtain the SVM classification model, and use the classification model to classify the user Classify the retrieved information to obtain the category of the user retrieved information;

[0068] The concrete steps of described step Step1 are:

[0069] Step1.1. Input the retrieved information in the knowledge base, perform word segmentation and filter stop word processing on the retrieved information, map each filtered word into a word vector, and then add the word vectors to form the retrieved information the sentence vector;

[0070] The present invention considers ...

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 relates to a cross-domain semantic information retrieval method based on a convolutional neural network, which belongs to the field of computer natural language processing. The present invention classifies short texts through word vector SVM, reduces the invalid retrieval domain and improves the accuracy of approximate sentences, and then splices the classified texts into a vector matrix and puts them into the convolutional neural network, and uses the last layer of the convolutional neural network to The conversion layer performs retrieval calculation of similar sentences. The final model improves the accuracy of approximate semantic retrieval.

Description

technical field [0001] The invention relates to a cross-domain semantic information retrieval method based on a convolutional neural network, which belongs to the field of computer natural language processing. Background technique [0002] The current keyword-based retrieval method has been widely recognized, but the hit rate is low. The reasons for the low hit rate include that keyword retrieval based on search engines requires a large amount of corpus, but in many scenarios, it is difficult to achieve the desired effect with small-scale or medium-scale corpus, and Zhao (<Chinese Journal of Computers>, 2005) proposed The keyword matching calculation method regards words as isolated elements, and it is unreasonable to have no connection with each other. At present, Zhao (<The Eighth National Joint Academic Conference on Computational Linguistics>, 2005) summarizes the research methods of sentence similarity as: 1) Sentence similarity calculation based on word fe...

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): G06F16/242G06K9/62G06N3/04
CPCG06F16/243G06N3/045G06F18/23G06F18/22G06F18/2411
Inventor 黄青松王兆凯李帅彬刘利军冯旭鹏
Owner KUNMING UNIV OF SCI & TECH
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