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

Rapid relation extraction method based on convolutional neural network and improved cascade labeling

A technology of convolutional neural network and relational extraction, which is applied in the field of fast relational extraction based on convolutional neural network and improved cascade labeling, can solve the problems of slow model training and prediction, and achieve good generalization, stability, and good The effect of speed and performance, good application prospect and promotion range

Active Publication Date: 2022-05-27
SOUTHEAST UNIV
View PDF3 Cites 2 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

First, a text encoder based on a deep neural network composed of dilated convolutions, gating units, and residual connections encodes the initial text to obtain a text encoding representation with rich contextual semantics, thus solving the problem of using a pre-trained model in relation extraction. As an encoder, the efficiency bottleneck problem that leads to slow model training and prediction

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
  • Rapid relation extraction method based on convolutional neural network and improved cascade labeling
  • Rapid relation extraction method based on convolutional neural network and improved cascade labeling
  • Rapid relation extraction method based on convolutional neural network and improved cascade labeling

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0116] As a specific embodiment of the present invention, the present invention provides training and prediction of a fast relation extraction method based on convolutional neural network and improved cascade annotation as shown in the following table for the training configuration flow table.

[0117]

[0118] The fast relation extraction model is used to train and predict relation extraction tasks on real Chinese and English data sets, and all other embodiments use the same data as this embodiment. Among them, the Chinese data set uses the DuIE relation extraction training set released by Baidu, which contains 11958 training data, 1498 verification data and 1489 test data, and defines 48 different relationship types; the English data set uses the New York Times corpus and The NYT relation extraction dataset obtained by Freebase's remote supervision contains 70,339 training data and 4,006 test data, and defines 29 different relation types. Due to the good robustness and ge...

Embodiment 2

[0121] The fast relation extraction model based on convolutional neural network and improved cascade annotation has fast reasoning ability. First of all, the model has less parameters, which is about 1% of the parameters of other relation extraction models at present, so the model requires calculation The time and computing resources are greatly reduced; when extracting relational triples in real-world natural text, the model takes less than 5ms on average to process a natural text, and the speed is 7-15 times faster than other methods.

[0122] In practical applications, it is often required that the model can process many natural texts at a time to ensure the parallelism of model operations. In the case of limited computing resources, the traditional relation extraction method using pre-trained language model as an encoder can only process and extract less than 8 or even 4 natural texts at a time, while the fast relation extraction model is limited by the amount of parameters...

Embodiment 3

[0124] The fast relation extraction model based on convolutional neural network and improved cascade annotation is easy to train and takes less time to train. Compared with other relation extraction methods, it has shorter turnaround time for single parameter update and less total parameters Update turnaround times so less total training time is required. For example, when the amount of training data is about 70,000, the batch size is set to 32, and the model is trained for 60 epochs, and it only takes about 100 minutes in the end, which is 3 to 10 times faster than other methods.

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 a fast relation extraction method based on a convolutional neural network and improved cascade labeling. The method comprises the following steps: firstly, encoding an initial text by a text encoder based on a deep neural network formed by connecting expansion convolution, a gating unit and residual errors to obtain a text encoding representation with rich context semantics; then, according to the obtained text codes, marking spans of all head entities and corresponding entity types by adopting improved cascade marking and a head entity marking device; then, through text coding representation and feature representation of the head entities, a tail entity annotator annotates all tail entities corresponding to each head entity; and finally, verifying through a relation extraction task in the real world. The method has the advantages of rapid training and prediction, and can meet the requirements of relation extraction scenes oriented to massive texts.

Description

technical field [0001] The invention belongs to the field of artificial intelligence natural language processing, in particular to a fast relation extraction method based on a convolutional neural network and an improved cascade annotation. Background technique [0002] Natural language processing technology is one of the core technologies of cognitive intelligence. It is the knowledge base that supports artificial intelligence application scenarios such as semantic search, intelligent question answering, and auxiliary big data analysis. The goal of the relation extraction task is to determine two entities in natural text. The relationship between. Recent research on relation extraction has made significant progress, however, in real application scenarios, relation extraction models need to maintain high demands on both speed and performance. For example, the investment decisions of investors in the financial sector rely on knowledge graphs consisting of relational triples ...

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 Applications(China)
IPC IPC(8): G06F40/279G06F40/126G06N3/04G06N3/08
CPCG06F40/279G06F40/126G06N3/08G06N3/045
Inventor 汪鹏李国正
Owner SOUTHEAST 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