Offensive Sentence Classification Using Character-Level CNN and Transfer Learning with Fake Sentences

Suin Seo; Sung Bea Cho

doi:10.1007/978-3-319-70096-0_55

Offensive Sentence Classification Using Character-Level CNN and Transfer Learning with Fake Sentences

Suin Seo, Sung Bea Cho

College of Computing

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution

5 Citations (Scopus)

Abstract

There are two difficulties in classifying offensive sentences: One is the modifiability of offensive terms, and the other is the class imbalance which appears in general offensive corpus. Solving these problems, we propose a method of pre-training fake sentences generated as character-level to convolution layers preventing under-fitting from data shortage, and dealing with the data imbalance. We insert the offensive words to half of the randomly generated sentences, and train the convolution neural networks (CNN) with theses sentences and the labels of whether offensive word is included. We use the trained filter of CNN for training new CNN given original data, resulting in the increase of the amount of training data. We get higher F1-score with the proposed method than that without pre-training in three dataset of insult from kaggle, Bullying trace, and formspring.

Original language	English
Title of host publication	Neural Information Processing - 24th International Conference, ICONIP 2017, Proceedings
Editors	Dongbin Zhao, El-Sayed M. El-Alfy, Derong Liu, Shengli Xie, Yuanqing Li
Publisher	Springer Verlag
Pages	532-539
Number of pages	8
ISBN (Print)	9783319700953
DOIs	https://doi.org/10.1007/978-3-319-70096-0_55
Publication status	Published - 2017
Event	24th International Conference on Neural Information Processing, ICONIP 2017 - Guangzhou, China Duration: 2017 Nov 14 → 2017 Nov 18

Publication series

Name	Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume	10635 LNCS
ISSN (Print)	0302-9743
ISSN (Electronic)	1611-3349

Other

Other	24th International Conference on Neural Information Processing, ICONIP 2017
Country/Territory	China
City	Guangzhou
Period	17/11/14 → 17/11/18

Bibliographical note

Funding Information:
Acknowledgments. This work was supported by Institute for Information & Communications Technology Promotion (IITP) grant funded by the Korea government (MSIP) (R0124-16-0002), Emotional Intelligence Technology to Infer Human Emotion and Carry on Dialogue Accordingly.

Publisher Copyright:
© 2017, Springer International Publishing AG.

All Science Journal Classification (ASJC) codes

Theoretical Computer Science
Computer Science(all)

Access to Document

10.1007/978-3-319-70096-0_55

Cite this

Seo, S., & Cho, S. B. (2017). Offensive Sentence Classification Using Character-Level CNN and Transfer Learning with Fake Sentences. In D. Zhao, E-S. M. El-Alfy, D. Liu, S. Xie, & Y. Li (Eds.), Neural Information Processing - 24th International Conference, ICONIP 2017, Proceedings (pp. 532-539). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 10635 LNCS). Springer Verlag. https://doi.org/10.1007/978-3-319-70096-0_55

Seo, Suin ; Cho, Sung Bea. / Offensive Sentence Classification Using Character-Level CNN and Transfer Learning with Fake Sentences. Neural Information Processing - 24th International Conference, ICONIP 2017, Proceedings. editor / Dongbin Zhao ; El-Sayed M. El-Alfy ; Derong Liu ; Shengli Xie ; Yuanqing Li. Springer Verlag, 2017. pp. 532-539 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).

@inproceedings{4010ff5a9f1746979766f9632da83e25,

title = "Offensive Sentence Classification Using Character-Level CNN and Transfer Learning with Fake Sentences",

abstract = "There are two difficulties in classifying offensive sentences: One is the modifiability of offensive terms, and the other is the class imbalance which appears in general offensive corpus. Solving these problems, we propose a method of pre-training fake sentences generated as character-level to convolution layers preventing under-fitting from data shortage, and dealing with the data imbalance. We insert the offensive words to half of the randomly generated sentences, and train the convolution neural networks (CNN) with theses sentences and the labels of whether offensive word is included. We use the trained filter of CNN for training new CNN given original data, resulting in the increase of the amount of training data. We get higher F1-score with the proposed method than that without pre-training in three dataset of insult from kaggle, Bullying trace, and formspring.",

author = "Suin Seo and Cho, {Sung Bea}",

note = "Funding Information: Acknowledgments. This work was supported by Institute for Information & Communications Technology Promotion (IITP) grant funded by the Korea government (MSIP) (R0124-16-0002), Emotional Intelligence Technology to Infer Human Emotion and Carry on Dialogue Accordingly. Publisher Copyright: {\textcopyright} 2017, Springer International Publishing AG.; 24th International Conference on Neural Information Processing, ICONIP 2017 ; Conference date: 14-11-2017 Through 18-11-2017",

year = "2017",

doi = "10.1007/978-3-319-70096-0_55",

language = "English",

isbn = "9783319700953",

series = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",

publisher = "Springer Verlag",

pages = "532--539",

editor = "Dongbin Zhao and El-Alfy, {El-Sayed M.} and Derong Liu and Shengli Xie and Yuanqing Li",

booktitle = "Neural Information Processing - 24th International Conference, ICONIP 2017, Proceedings",

address = "Germany",

}

Seo, S & Cho, SB 2017, Offensive Sentence Classification Using Character-Level CNN and Transfer Learning with Fake Sentences. in D Zhao, E-SM El-Alfy, D Liu, S Xie & Y Li (eds), Neural Information Processing - 24th International Conference, ICONIP 2017, Proceedings. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 10635 LNCS, Springer Verlag, pp. 532-539, 24th International Conference on Neural Information Processing, ICONIP 2017, Guangzhou, China, 17/11/14. https://doi.org/10.1007/978-3-319-70096-0_55

Offensive Sentence Classification Using Character-Level CNN and Transfer Learning with Fake Sentences. / Seo, Suin; Cho, Sung Bea.

Neural Information Processing - 24th International Conference, ICONIP 2017, Proceedings. ed. / Dongbin Zhao; El-Sayed M. El-Alfy; Derong Liu; Shengli Xie; Yuanqing Li. Springer Verlag, 2017. p. 532-539 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 10635 LNCS).

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution

TY - GEN

T1 - Offensive Sentence Classification Using Character-Level CNN and Transfer Learning with Fake Sentences

AU - Seo, Suin

AU - Cho, Sung Bea

N1 - Funding Information: Acknowledgments. This work was supported by Institute for Information & Communications Technology Promotion (IITP) grant funded by the Korea government (MSIP) (R0124-16-0002), Emotional Intelligence Technology to Infer Human Emotion and Carry on Dialogue Accordingly. Publisher Copyright: © 2017, Springer International Publishing AG.

PY - 2017

Y1 - 2017

N2 - There are two difficulties in classifying offensive sentences: One is the modifiability of offensive terms, and the other is the class imbalance which appears in general offensive corpus. Solving these problems, we propose a method of pre-training fake sentences generated as character-level to convolution layers preventing under-fitting from data shortage, and dealing with the data imbalance. We insert the offensive words to half of the randomly generated sentences, and train the convolution neural networks (CNN) with theses sentences and the labels of whether offensive word is included. We use the trained filter of CNN for training new CNN given original data, resulting in the increase of the amount of training data. We get higher F1-score with the proposed method than that without pre-training in three dataset of insult from kaggle, Bullying trace, and formspring.

AB - There are two difficulties in classifying offensive sentences: One is the modifiability of offensive terms, and the other is the class imbalance which appears in general offensive corpus. Solving these problems, we propose a method of pre-training fake sentences generated as character-level to convolution layers preventing under-fitting from data shortage, and dealing with the data imbalance. We insert the offensive words to half of the randomly generated sentences, and train the convolution neural networks (CNN) with theses sentences and the labels of whether offensive word is included. We use the trained filter of CNN for training new CNN given original data, resulting in the increase of the amount of training data. We get higher F1-score with the proposed method than that without pre-training in three dataset of insult from kaggle, Bullying trace, and formspring.

UR - http://www.scopus.com/inward/record.url?scp=85035127324&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85035127324&partnerID=8YFLogxK

U2 - 10.1007/978-3-319-70096-0_55

DO - 10.1007/978-3-319-70096-0_55

M3 - Conference contribution

AN - SCOPUS:85035127324

SN - 9783319700953

T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)

SP - 532

EP - 539

BT - Neural Information Processing - 24th International Conference, ICONIP 2017, Proceedings

A2 - Zhao, Dongbin

A2 - El-Alfy, El-Sayed M.

A2 - Liu, Derong

A2 - Xie, Shengli

A2 - Li, Yuanqing

PB - Springer Verlag

T2 - 24th International Conference on Neural Information Processing, ICONIP 2017

Y2 - 14 November 2017 through 18 November 2017

ER -

Seo S, Cho SB. Offensive Sentence Classification Using Character-Level CNN and Transfer Learning with Fake Sentences. In Zhao D, El-Alfy E-SM, Liu D, Xie S, Li Y, editors, Neural Information Processing - 24th International Conference, ICONIP 2017, Proceedings. Springer Verlag. 2017. p. 532-539. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)). doi: 10.1007/978-3-319-70096-0_55

Offensive Sentence Classification Using Character-Level CNN and Transfer Learning with Fake Sentences

Abstract

Publication series

Other

Bibliographical note

All Science Journal Classification (ASJC) codes

Access to Document

Other files and links

Fingerprint

Cite this