Address information feature extraction method based on deep neural network model

A deep neural network and address information technology, applied in the field of address information feature extraction based on deep neural network model, can solve the problems of inability to deeply mine the connotation of text features, additional data dependencies, and information islands.

Active Publication Date: 2019-10-25
ZHEJIANG UNIV
View PDF4 Cites 48 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

However, the current theoretical research centered on unstructured text management or address coding is unable to dig deep into the characteristic connotation of text, leading to p

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
  • Address information feature extraction method based on deep neural network model
  • Address information feature extraction method based on deep neural network model
  • Address information feature extraction method based on deep neural network model

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0250] In this embodiment, an address text data set is constructed with 2 million pieces of place name and address data in Shangcheng District, Hangzhou City, and feature vector extraction is performed on it. The basic steps are as described in the aforementioned S1-S7, and will not be described in detail. The following mainly demonstrates some specific implementation details and effects of each step.

[0251] 1. According to the method described in steps S1-S7, use the TensorFlow deep learning framework to build ALM and GSAM, and set the save point of the model at the same time, save the neural network parameter variables except the target task module, so as to facilitate the transplantation in the next fine-tuning task; The hyperparameters of the model are set through the hype-para.config configuration file, and the specific contents mainly include the following categories:

[0252] 1) Training sample size batch_size: 64; 2) Initial learning rate η: 0.00005; 3) Number of tra...

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 address information feature extraction method based on a deep neural network model. According to the method, tasks such as text feature extraction, address standardization construction and semantic space fusion are converted into quantifiable deep neural network model construction and training optimization problems by utilizing a deep neural network architecture. Characters in an address are used as basic input units, a language model is designed to express the characters in a vectorization mode, and then the key technology of place name and address standardized construction is achieved through a neural network target task. Meanwhile, considering place name address space expression characteristics, proposing an address semantic-space characteristic fusion scheme,designing a weighted clustering method and a characteristic fusion model, and extracting a fusion vector fused with semantic characteristics and space characteristics from an address text of a natural language. The feature content extraction of the address information can be realized. The structure has high expansibility. The solution thinking of address information tasks can be unified. The method is of great significance to urban construction.

Description

technical field [0001] The invention relates to the field of address information mining of GIS (Geographic Information System), in particular to a method for extracting features of address information based on a deep neural network model. Background technique [0002] With the continuous improvement of GIS cognition and application capabilities, address information has gradually become the core resource in the era of smart cities. The semantic and spatial connotations carried in its content are the basic support for the construction of geographic ontology and spatiotemporal semantic framework in smart cities. It is of great theoretical and practical significance for the integration and understanding of urban semantics and spatial content to allow computers to deeply refine the comprehensive characteristics of place names and addresses from the perspective of understanding address texts and form quantitative expressions in numerical form. However, the current theoretical rese...

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): G06F16/29G06F16/35G06F17/27G06F17/22G06K9/62
CPCG06F16/29G06F16/35G06F40/126G06F40/30G06F18/23213G06N3/084G06N3/048G06N3/045G06N3/08G06N7/01
Inventor 张丰毛瑞琛杜震洪徐流畅叶华鑫
Owner ZHEJIANG 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