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

Method and system for automatic treebank conversion based on tree-shaped recurrent neural network

A cyclic neural network and tree technology, applied in neural learning methods, biological neural network models, neural architectures, etc., can solve problems such as general transformation effect, insufficient utilization of source tree bank, and missing dual-tree alignment data.

Active Publication Date: 2022-03-15
SUZHOU UNIV
View PDF3 Cites 0 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

The main problem of the indirect method is that the source-end treebank is not fully utilized, and it cannot effectively describe the corresponding rules between norms; while the direct method based on transformation is limited by the lack of double-tree alignment data, and cannot effectively learn the relationship between norms. Corresponding law, so the conversion effect is average

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 and system for automatic treebank conversion based on tree-shaped recurrent neural network
  • Method and system for automatic treebank conversion based on tree-shaped recurrent neural network
  • Method and system for automatic treebank conversion based on tree-shaped recurrent neural network

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0071] The present embodiment is based on the automatic tree bank conversion method of the tree-shaped recurrent neural network, including:

[0072] Obtaining a double-tree alignment database, which stores sentences marked with two annotation specifications;

[0073] Calculate respectively the arc-score value of the dependence of every two words in the target terminal tree in each described sentence, wherein, the two words described are separated by word w i and the word w j Indicates that the presupposition word w i and the word w j In the target tree, they are respectively modifiers and core words, word w i and the word w j The calculation process of dependent arc minutes in the target tree includes:

[0074] Extract the word w in the source tree i , word w j The shortest path tree of , based on the bidirectional tree-shaped recurrent neural network TreeLSTM, the word w in the shortest path tree is obtained i , word w j , word w a The respective hidden layer output...

example 1

[0110] Example 1: see image 3 .1, the shortest path tree of word C, word A, and word B, where word A corresponds to word w i , word B corresponds to word w j , word C corresponds to word w a .

[0111] Operation from the bottom up:

[0112] (1) Calculate the hidden layer output vector of word A: part of the input information of the LSTM node is: the top layer output vector corresponding to word A; the other part is the zero vector.

[0113] (2) Calculate the hidden layer output vector of word B: part of the input information of the LSTM node is: the top layer output vector corresponding to word B; the other part is the zero vector.

[0114] (3) Calculate the hidden layer output vector of the word C as the ancestor node: part of the input information of the LSTM node is: the top layer output vector corresponding to the word C; since the word C has two sons, the other part is the child node word A and word B's hidden layer output vector. so far image 3 .1 All calculatio...

example 2

[0119] Example 2: see image 3 .2, the shortest path tree of word E, word C, and word D, where word D corresponds to word w i , word C corresponds to word w j , word E corresponds to word w a . Word E is the closest common ancestor node of word C and word D, and the calculation method is the same as that of Example 1, and will not be repeated here.

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 an automatic tree bank conversion method and system based on a tree-shaped recurrent neural network, which is designed to obtain an accurate supervised conversion model. The automatic tree bank conversion method based on the tree-shaped recurrent neural network of the present invention includes: based on the bidirectional tree-shaped recurrent neural network TreeLSTM, the word w is obtained i , word w i , word w a The hidden layer output vector of the hidden layer output vector is spliced ​​together as the word w i and the word w j In the source tree, the representation vector output vector of the top layer of the recurrent neural network BiSeqLSTM is concatenated with the representation vector respectively, and used as the input of the perceptron MLP, the perceptron extracts syntactic information; the word w is calculated using biaffine i and the word w j The target side of the depends on the arc minutes value. The invention makes full use of the source-end syntax tree to describe the corresponding rules of the two labeling norms and provides necessary data support for establishing a high-quality tree supervised transformation model.

Description

technical field [0001] The invention belongs to the technical field of natural language processing, and in particular relates to an automatic tree bank conversion method and system based on a tree-shaped recurrent neural network. Background technique [0002] At present, researchers have carried out a lot of research and development work in tree bank research, and have also achieved considerable results. The annotation systems used by these treebanks are very different, and they can be roughly divided into two types according to the description method, one is the phrase structure tree, and the other is the dependency tree. [0003] For the dependency tree, the dependency annotations of the two treebanks follow different annotation specifications, and the two treebanks are said to be heterogeneous. Many mainstream languages ​​in the world have multiple large-scale heterogeneous treebanks. Since the construction of treebanks requires a very high labor cost, how to use differe...

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): G06F40/211G06F40/289G06N3/04G06N3/08
CPCG06N3/08G06F40/211G06F40/289G06N3/045
Inventor 李正华江心舟章波张民陈文亮
Owner SUZHOU 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