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

Text classification algorithm based on bidirectional dynamic routing with label constraint

A text classification and labeling technology, applied in text database clustering/classification, text database query, unstructured text data retrieval, etc., can solve problems such as unreasonable distribution of capsule networks, and achieve the effect of solving unreasonable distribution

Pending Publication Date: 2021-12-10
HOHAI UNIV
View PDF0 Cites 0 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0004] Purpose of the invention: In order to solve the problems in the above-mentioned background technology, the present invention proposes a text classification algorithm based on two-way dynamic routing with label constraints to solve the problem of potential unreasonable distribution of capsule networks

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
  • Text classification algorithm based on bidirectional dynamic routing with label constraint
  • Text classification algorithm based on bidirectional dynamic routing with label constraint
  • Text classification algorithm based on bidirectional dynamic routing with label constraint

Examples

Experimental program
Comparison scheme
Effect test

Embodiment Construction

[0071] The present invention will be further described below in conjunction with the accompanying drawings.

[0072] Such as figure 1 A text classification algorithm based on two-way dynamic routing with label constraints is shown, including the following steps:

[0073] Step S1, Embedding Layer: Encode the given text, each word vector and label vector are randomly initialized to initialize the distribution; specifically,

[0074] Given a sequence of words X = {w 1 ,w 2 …w M}, where M represents the number of words in the text, w i Represents the word at the i-th position in the text; using the word embedding method: embedding each word in the sequence into a high-dimensional vector space, as follows:

[0075] E. X ={e 1 , e 2 ...e M}

[0076] where word embedding matrix where V is the size of the dictionary and D w is the dimension of the word vector.

[0077] Step S2, BiLSTM Layer: In the word vector, each word is independent of other words in the sentence. For...

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 provides a text classification algorithm based on bidirectional dynamic routing with label constraints, which comprises the following steps of: firstly, fusing semantic information and word order information into the same vector through a self-attention mechanism and relative position information coding; then introducing a bidirectional dynamic routing mechanism, and simultaneously considering the conversion amount information from the bottom layer capsule to the high layer capsule and the amount information of the bottom layer capsule required by the high layer capsule; finally, a label embedding constraint method is provided to replace model reconstruction in a capsule network, and the potential unreasonable distribution condition of text capsules is solved; The relative position coding is introduced, the problem that a dynamic routing algorithm neglects the word order of words is solved, and the bidirectional dynamic routing algorithm can give consideration to the problem that the same word has different importance in different types of texts; the provided label embedding constraint method can solve the problem of potential unreasonable distribution of text capsules.

Description

technical field [0001] The invention relates to the technical field of text classification, and mainly relates to a text classification algorithm based on bidirectional dynamic routing with label constraints. Background technique [0002] Because text classification plays an important role in natural language, it has become a relatively active research field in recent years. It has achieved great success in inventions such as document retrieval, web search, and email filtering. Moreover, many natural language processing Tasks can eventually be transformed into text classification tasks, such as sentiment analysis and intent detection. The text classification task is to annotate a given sequence of text with one (or more) class labels. How to construct an effective text feature representation is the most critical step in the text classification task. Existing methods for text classification can be mainly divided into two categories: 1) Traditional methods based on bag-of-wo...

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): G06F16/33G06F16/35G06F40/284G06F40/30G06K9/62G06N3/04G06N3/08
CPCG06F16/3344G06F16/35G06F40/284G06F40/30G06N3/08G06N3/044G06F18/241
Inventor 曹杰王有权陈雷郭翔申冬琴
Owner HOHAI UNIV
Features
  • Generate Ideas
  • 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