CNN MODELS FOR READABILITY OF CHINESE TEXTS

Han FENG; Sizai HOU; Le-Yin WEI; Ding-Xuan ZHOU

doi:10.3934/mfc.2022021

CNN MODELS FOR READABILITY OF CHINESE TEXTS

Han FENG, Sizai HOU, Le-Yin WEI^*, Ding-Xuan ZHOU

^*Corresponding author for this work

Research output: Journal Publications and Reviews › RGC 21 - Publication in refereed journal › peer-review

28 Citations (Scopus)

Abstract

Readability of Chinese texts considered in this paper is a multi-class classification problem with 12 grade classes corresponding to 6 grades in primary schools, 3 grades in middle schools, and 3 grades in high schools. A special property of this problem is the strong ambiguity in determining the grades. To overcome the difficulty, a measurement of readability assessment methods used empirically in practice is adjacent accuracy in addition to exact accuracy. In this paper we give mathematical definitions of these concepts in a learning theory framework and compare these two quantities in terms of the ambiguity level of texts. A deep learning algorithm is proposed for readability of Chinese texts, based on convolutional neural networks and a pre-trained BERT model for vector representations of Chinese characters. The proposed CNN model can extract sentence and text features by convolutions of sentence representations with filters and is efficient for readability assessment, which is demonstrated with some numerical experiments.

Original language	English
Pages (from-to)	351-362
Journal	Mathematical Foundations of Computing
Volume	5
Issue number	4
Online published	Jul 2022
DOIs	https://doi.org/10.3934/mfc.2022021
Publication status	Published - Nov 2022

Funding

The first author is supported partially by the Research Grants Council of Hong Kong [Projects # CityU 11303821, # CityU 11306220]. The third author is supported partially by National Science Foundation of China under the NSFC/RGC Joint Research Scheme [Project No. 12061160462]. The last author is supported partially by the Research Grants Council of Hong Kong [Project # C1013-21GF], the NSFC/RGC Joint Research Scheme [RGC Project No. N CityU102/20 and NSFC Project No. 12061160462], the Germany/Hong Kong Joint Research Scheme [Project No. G-CityU101/20], the CityU Strategic Interdisciplinary Research Grant [Project No. 7020010], Hong Kong Institute for Data Science and the Laboratory for AI-Powered Financial Technologies. We would like to thank Professor Jinshan Zeng for his suggestions.

Research Keywords

Adjacent accuracy
Cnn
Filters
Neural networks
Readability of Chinese texts

Access Output

10.3934/mfc.2022021

Other Access Options

Check CityUHK Library

Permanent Link

Right click to copy link

Cite this

@article{ff128c355b014232b52e38a17c45f7a0,

title = "CNN MODELS FOR READABILITY OF CHINESE TEXTS",

abstract = "Readability of Chinese texts considered in this paper is a multi-class classification problem with 12 grade classes corresponding to 6 grades in primary schools, 3 grades in middle schools, and 3 grades in high schools. A special property of this problem is the strong ambiguity in determining the grades. To overcome the difficulty, a measurement of readability assessment methods used empirically in practice is adjacent accuracy in addition to exact accuracy. In this paper we give mathematical definitions of these concepts in a learning theory framework and compare these two quantities in terms of the ambiguity level of texts. A deep learning algorithm is proposed for readability of Chinese texts, based on convolutional neural networks and a pre-trained BERT model for vector representations of Chinese characters. The proposed CNN model can extract sentence and text features by convolutions of sentence representations with filters and is efficient for readability assessment, which is demonstrated with some numerical experiments.",

keywords = "Adjacent accuracy, Cnn, Filters, Neural networks, Readability of Chinese texts",

author = "Han FENG and Sizai HOU and Le-Yin WEI and Ding-Xuan ZHOU",

year = "2022",

month = nov,

doi = "10.3934/mfc.2022021",

language = "English",

volume = "5",

pages = "351--362",

number = "4",

}

TY - JOUR

T1 - CNN MODELS FOR READABILITY OF CHINESE TEXTS

AU - FENG, Han

AU - HOU, Sizai

AU - WEI, Le-Yin

AU - ZHOU, Ding-Xuan

PY - 2022/11

Y1 - 2022/11

N2 - Readability of Chinese texts considered in this paper is a multi-class classification problem with 12 grade classes corresponding to 6 grades in primary schools, 3 grades in middle schools, and 3 grades in high schools. A special property of this problem is the strong ambiguity in determining the grades. To overcome the difficulty, a measurement of readability assessment methods used empirically in practice is adjacent accuracy in addition to exact accuracy. In this paper we give mathematical definitions of these concepts in a learning theory framework and compare these two quantities in terms of the ambiguity level of texts. A deep learning algorithm is proposed for readability of Chinese texts, based on convolutional neural networks and a pre-trained BERT model for vector representations of Chinese characters. The proposed CNN model can extract sentence and text features by convolutions of sentence representations with filters and is efficient for readability assessment, which is demonstrated with some numerical experiments.

AB - Readability of Chinese texts considered in this paper is a multi-class classification problem with 12 grade classes corresponding to 6 grades in primary schools, 3 grades in middle schools, and 3 grades in high schools. A special property of this problem is the strong ambiguity in determining the grades. To overcome the difficulty, a measurement of readability assessment methods used empirically in practice is adjacent accuracy in addition to exact accuracy. In this paper we give mathematical definitions of these concepts in a learning theory framework and compare these two quantities in terms of the ambiguity level of texts. A deep learning algorithm is proposed for readability of Chinese texts, based on convolutional neural networks and a pre-trained BERT model for vector representations of Chinese characters. The proposed CNN model can extract sentence and text features by convolutions of sentence representations with filters and is efficient for readability assessment, which is demonstrated with some numerical experiments.

KW - Adjacent accuracy

KW - Cnn

KW - Filters

KW - Neural networks

KW - Readability of Chinese texts

UR - http://www.scopus.com/inward/record.url?scp=85136647821&partnerID=8YFLogxK

UR - http://gateway.isiknowledge.com/gateway/Gateway.cgi?GWVersion=2&SrcAuth=LinksAMR&SrcApp=PARTNER_APP&DestLinkType=FullRecord&DestApp=WOS&KeyUT=000826430600001

UR - https://www.scopus.com/record/pubmetrics.uri?eid=2-s2.0-85136647821&origin=recordpage

U2 - 10.3934/mfc.2022021

DO - 10.3934/mfc.2022021

M3 - RGC 21 - Publication in refereed journal

SN - 2577-8838

VL - 5

SP - 351

EP - 362

JO - Mathematical Foundations of Computing

JF - Mathematical Foundations of Computing

IS - 4

ER -

CNN MODELS FOR READABILITY OF CHINESE TEXTS

Abstract

Funding

Research Keywords

Access Output

Other Access Options

Permanent Link

Other Files and Links

Fingerprint

GRF: Approximation Analysis of Classification with Deep Convolutional Neural Networks

EU_RGC: Development of One-step Food Waste Biorefinery via Novel Bioreactor Design, Functional Strain Adaptive Laboratory Evolution and Genetic Engineering

Ger/HKJRS: Theoretical Research on Deep Learning from a Mathematical Approximation Theory Viewpoint

Cite this