Article title detection method based on a hierarchical hybrid network and a federated learning strategy

A hybrid network, title party technology, applied in the field of machine learning, can solve problems such as not using information

Pending Publication Date: 2019-04-19
SUN YAT SEN UNIV
View PDF3 Cites 13 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

However, although the document classification method based on the bidirectional recurrent neural network of the multi-layer attention mechanism performs excellently on most document classification tasks, it does not utiliz...

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
  • Article title detection method based on a hierarchical hybrid network and a federated learning strategy
  • Article title detection method based on a hierarchical hybrid network and a federated learning strategy
  • Article title detection method based on a hierarchical hybrid network and a federated learning strategy

Examples

Experimental program
Comparison scheme
Effect test

no. 1 example

[0063] In a specific embodiment of the present invention, the following scenario is considered: there are two content platforms A and B. Platform A is a content platform for disseminating traditional articles. Assume that platform A has marked part of the data (whether it belongs to the headline party), the data has title t, article content b, and the label l obtained after marking, that is, the data Xa of platform A has always been [t , b], platform A uses labeled data to train a headline party detection model of the aforementioned hierarchical hybrid network, that is, through the network Wa (the network Wa can be considered as the headline encoder part, content encoder part, and association part of the hierarchical hybrid network The set of the information extractor part) to obtain the feature vector Va, and Va is obtained by Vt (the feature vector of the title obtained through the title encoder of the hierarchical hybrid network), Vb (the feature vector of the content obtain...

no. 2 example

[0085] Similarly, in a specific embodiment of the present invention, the following scenarios are considered: Assume that there are two content platforms A and B, and platform A is a content platform for disseminating traditional articles. The title t, the content of the article, and the label l obtained by marking it out, that is, the data Xa of platform A is always [t,b]; platform B is a platform like Weibo, and the data has title t, abstract description d, and picture i , the keyword k, that is, the vector Xb of platform B is always [t; d; i; k].

[0086] Specifically, platform A and platform B want to cooperate (platform B has no labels) to jointly train a model C that is more effective (compared to model A). However, due to data privacy protection regulations, the two parties cannot share their own data. At this time, platform A and platform B want to cooperate (platform B has no label), and jointly train a model C with better effect (compared to model A). In this embodime...

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 article title detection method based on a hierarchical hybrid network and a federated learning strategy, and the model comprises a title encoder which is used for carrying out feature extraction on article titles and effectively encoding article title texts into title vectors; The content encoder is used for carrying out feature extraction on the content text and effectively encoding the content text into a document vector; The associated information extractor is used for associating the title vector with the document vector by using a machine reading understanding related technology so as to obtain an associated vector of the title vector and the document vector; And the classification network is used for carrying out title codonopsis pilosula classification based on the title feature vectors, the document vectors and the association vectors, and a better title codonopsis pilosula detection effect can be obtained by utilizing the association information between the document titles and the document contents.

Description

technical field [0001] The invention relates to the technical field of machine learning, in particular to a method for detecting headline party articles based on a hierarchical hybrid network and a federated learning strategy. Background technique [0002] At present, the income of most Internet media depends on the number of user clicks. Due to fierce competition, in order to attract users' attention, some media will use an exaggerated and eye-catching title when writing an article. However, when users are attracted by the title and click to watch the content of the article, they are often disappointed. This kind of article with a large gap between the title and the content is called the title party. [0003] In order to reduce the spread of headline partying, people began to study how to use machine learning technology to detect headline partying. A common practice in academia and industry is to treat headline party detection as a text classification problem. Most of t...

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): G06F16/35G06F17/27G06N3/04G06N3/08
CPCG06N3/08G06F40/258G06N3/045
Inventor 廖枫卓汉逵
Owner SUN YAT SEN UNIV
Who we serve
  • R&D Engineer
  • R&D Manager
  • IP Professional
Why Eureka
  • Industry Leading Data Capabilities
  • Powerful AI technology
  • Patent DNA Extraction
Social media
Try Eureka
PatSnap group products

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

© 2024 PatSnap. All rights reserved.Legal|Privacy policy|Modern Slavery Act Transparency Statement|Sitemap