Unified low sample relation extraction method and device based on multi-choice matching network

A technology of matching network and relation extraction, applied in neural learning methods, biological neural network models, unstructured text data retrieval, etc. Calculation cost and calculation speed, and the effect of improving model performance

Pending Publication Date: 2022-05-24
INST OF SOFTWARE - CHINESE ACAD OF SCI
View PDF0 Cites 0 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0007] In order to overcome the problem of high computational cost in the existing unified low-sample relation extraction method, the present invention proposes a unified relation extraction method and device based on a multi-choice matching network, and models the relation ex

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
  • Unified low sample relation extraction method and device based on multi-choice matching network
  • Unified low sample relation extraction method and device based on multi-choice matching network
  • Unified low sample relation extraction method and device based on multi-choice matching network

Examples

Experimental program
Comparison scheme
Effect test

Embodiment

[0066] For the relationship extraction task containing the following three relationship categories: "employee", "employer", "investor", the example to be classified is: "Cook is the CEO of Apple", and it is processed correspondingly, and Relation extraction is performed through a multi-choice matching network.

[0067] Implementation:

[0068] (1) A description of all target relationships of the current task. Concatenated into multiple-choice statements:

[0069] [choice] employee [choice] employer [choice] investor [choice] other

[0070] (2) Splicing the instance to be classified with the multiple-choice statement:

[0071] [choice]employee[choice]employer[choice]investor[choice]other[sep][e1]Cook[ / e1] is the CEO of [e2]Apple[ / e2]

[0072] (3) Input the above processed results into the multi-selection matching network, and calculate the similarity between the instance to be classified and each category, and the most similar relationship is the prediction result. In this...

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 unified low sample relation extraction method and device based on a multi-choice matching network. The method comprises the following steps: based on a pre-training language model and a multi-choice marked relationship description and relationship instance common coding and matching mechanism; the triple is obtained through extraction based on open information of a large-scale plain text, the retelling text is generated through a generative pre-training language model, and a triple-retelling pre-training mode is carried out based on the retelling text; the invention relates to an online element learning training mode based on small samples under a new task. According to the mechanism based on the multi-choice matching network, various scenes in a low-sample relation extraction task can be modeled in a unified manner, and an efficient and rapid network architecture is provided, so that the model better meets multiple requirements on model performance and speed in practical application.

Description

technical field [0001] The invention relates to a low-sample relationship extraction method, in particular to a low-sample relationship extraction method and device based on a multi-selection matching network and its pre-training and online training methods, belonging to the technical field of natural language processing. Background technique [0002] Relation extraction is an important task in information extraction and natural language understanding. The task aims to identify the class of relationship expressed by a pair of entities, given the context. E.g. Given the context "A is the founder of Company B", and the entity pair "A" and "Company B", a relation extraction system should be able to correctly identify the entity pair expressing the relation "is the founder of...". [0003] In recent years, with the development of deep learning methods, the accuracy of relation extraction systems has been greatly improved. However, such methods require a large amount of high-q...

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): G06F16/35G06F40/30G06K9/62G06N3/08
CPCG06F16/35G06F40/30G06N3/08G06F18/22G06F18/214
Inventor 刘方超林鸿宇韩先培孙乐
Owner INST OF SOFTWARE - CHINESE ACAD OF SCI
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