Unlock instant, AI-driven research and patent intelligence for your innovation.

A small-sample fine-grained entity classification method based on relational graph convolutional network

A technology of convolutional network and classification method, applied in the field of small-sample fine-grained entity classification based on relational graph convolutional network, can solve problems such as poor effect, and achieve the effect of improving robustness and high classification accuracy

Active Publication Date: 2022-05-24
ZHEJIANG UNIV
View PDF3 Cites 0 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

They tend to be less effective when labeled data is limited

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
  • A small-sample fine-grained entity classification method based on relational graph convolutional network
  • A small-sample fine-grained entity classification method based on relational graph convolutional network
  • A small-sample fine-grained entity classification method based on relational graph convolutional network

Examples

Experimental program
Comparison scheme
Effect test

Embodiment

[0123] Below in conjunction with the method of the present invention, the concrete steps that this embodiment is implemented is described in detail, as follows:

[0124] In this embodiment, the method of the present invention is applied to a common data set FIGER for fine-grained entity classification, 10 categories are randomly selected as small sample categories, and each category has K (K=5 or 10) labels. example, to classify other target entities of these few-shot categories.

[0125] 1) Divide the dataset for each episode. The FIGER dataset contains a total of 128 categories. After excluding 10 small sample categories, 118 categories are actually used for training. In each episode, imitating the setting of few-shot learning, 10 classes are randomly selected from 118 classes as few-shot classes, and K (K=5 or 10) samples are randomly selected for each class, thus forming the support set . The remaining 108 categories are used as frequent sample categories to constitute ...

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 small-sample fine-grained entity classification method based on a relational graph convolution network. First, the entity to be classified in the sentence is encoded by the "entity-context encoder", and the robustness of the model is enhanced by adversarial learning, and new data is automatically generated by data augmentation. Second, a category co-occurrence graph that can effectively perceive small samples is constructed, and categories are encoded using a relational graph convolutional network. Afterwards, entities are classified by matching the encodings of entities and categories. The entire model is trained via meta-learning. Finally, by inputting small-sample class data into the model to fine-tune its parameters, the model can be used to classify other target entities of these small-sample categories.

Description

technical field [0001] The invention belongs to the technical field of entity classification, and in particular relates to a small sample fine-grained entity classification method based on a relational graph convolution network. Background technique [0002] Fine-grained entity classification is a fundamental task in natural language processing, which is to assign appropriate categories to specific entities that contain surrounding context. The concept of category is the most basic unit of constructing the thinking system when human beings know the world. With categories, one can extract the common essential characteristics of entities under the category without tangling with the nuances. In addition, the correct entity category information will also help people better understand and recognize new entities, which can also serve more downstream tasks. [0003] A large number of deep learning models composed of convolutional neural networks have been proposed to solve this t...

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): G06F16/35G06F40/126G06F40/295G06K9/62G06N3/04G06N3/08
CPCG06F16/35G06F40/126G06F40/295G06N3/084G06N3/044G06N3/045G06F18/241G06F18/214
Inventor 鲁伟明陈晨庄越挺
Owner ZHEJIANG UNIV