Bots and gender profiling of tweets using word and character N-grams notebook for PAN at CLEF 2019

Yaakov HaCohen-Kerner; Natan Manor; Michael Goldmeier

Bots and gender profiling of tweets using word and character N-grams notebook for PAN at CLEF 2019

Yaakov HaCohen-Kerner
, Natan Manor
, Michael Goldmeier

Jerusalem College of Technology

Research output: Contribution to journal › Conference article › peer-review

1 Scopus citations

Abstract

Author profiling deals with the identification of various details about the author of the text (e.g., age and gender). In this paper, we describe the participation of our team (hacohenkerner19) in the PAN 2019 shared task on Bots and Gender Profiling in two languages: English and Spanish. Given a Twitter feed, we should determine whether its author is a bot or a human. In the case of human, we should identify her/his gender. In this paper, we describe our preprocessing methods, feature sets, five applied machine learning methods, and accuracy results. The best accuracy result for the English dataset (84.8%) was obtained by LinearSVC using 2,000 word unigrams. The same result (84.8%) was also obtained by LR by using four preprocessing methods, 2,000 word unigrams, and 1,000 word bigrams with maximal skips of 2 words. The best accuracy result (75,54%) for the Spanish dataset was achieved using LinearSVC with only the HTML tag removal preprocessing method and a combination of 1,000 word unigrams, 1,000 word bigrams, and 3,000 character trigrams.

Original language	English
Journal	CEUR Workshop Proceedings
Volume	2380
State	Published - 2019
Externally published	Yes
Event	20th Working Notes of CLEF Conference and Labs of the Evaluation Forum, CLEF 2019 - Lugano, Switzerland Duration: 9 Sep 2019 → 12 Sep 2019

Bibliographical note

Publisher Copyright:
© 2019 for this paper by its authors. Use permitted under Creative Commons License Attribution 4.0 International (CC BY 4.0). CLEF 2019, 9-12 September 2019, Lugano, Switzerland.

Funding

Acknowledgments. This work was partially funded by the Jerusalem College Technology (Lev Academic Center) and we gratefully acknowledge its support. This work was partially funded by the Jerusalem College of Technology (Lev Academic Center) and we gratefully acknowledge its support.

Funders
Jerusalem College Technology
Jerusalem College of Technology

Keywords

Bot Profiling
Character N-grams
Gender Profiling
Supervised Machine Learning
Word N-grams

Cite this

@article{8dc5be090f4a4caa82e473c7c3ed6416,

title = "Bots and gender profiling of tweets using word and character N-grams notebook for PAN at CLEF 2019",

abstract = "Author profiling deals with the identification of various details about the author of the text (e.g., age and gender). In this paper, we describe the participation of our team (hacohenkerner19) in the PAN 2019 shared task on Bots and Gender Profiling in two languages: English and Spanish. Given a Twitter feed, we should determine whether its author is a bot or a human. In the case of human, we should identify her/his gender. In this paper, we describe our preprocessing methods, feature sets, five applied machine learning methods, and accuracy results. The best accuracy result for the English dataset (84.8\%) was obtained by LinearSVC using 2,000 word unigrams. The same result (84.8\%) was also obtained by LR by using four preprocessing methods, 2,000 word unigrams, and 1,000 word bigrams with maximal skips of 2 words. The best accuracy result (75,54\%) for the Spanish dataset was achieved using LinearSVC with only the HTML tag removal preprocessing method and a combination of 1,000 word unigrams, 1,000 word bigrams, and 3,000 character trigrams.",

keywords = "Bot Profiling, Character N-grams, Gender Profiling, Supervised Machine Learning, Word N-grams",

author = "Yaakov HaCohen-Kerner and Natan Manor and Michael Goldmeier",

note = "Publisher Copyright: {\textcopyright} 2019 for this paper by its authors. Use permitted under Creative Commons License Attribution 4.0 International (CC BY 4.0). CLEF 2019, 9-12 September 2019, Lugano, Switzerland.; 20th Working Notes of CLEF Conference and Labs of the Evaluation Forum, CLEF 2019 ; Conference date: 09-09-2019 Through 12-09-2019",

year = "2019",

language = "אנגלית",

volume = "2380",

journal = "CEUR Workshop Proceedings",

issn = "1613-0073",

publisher = "CEUR-WS",

}

TY - JOUR

T1 - Bots and gender profiling of tweets using word and character N-grams notebook for PAN at CLEF 2019

AU - HaCohen-Kerner, Yaakov

AU - Manor, Natan

AU - Goldmeier, Michael

N1 - Publisher Copyright: © 2019 for this paper by its authors. Use permitted under Creative Commons License Attribution 4.0 International (CC BY 4.0). CLEF 2019, 9-12 September 2019, Lugano, Switzerland.

PY - 2019

Y1 - 2019

N2 - Author profiling deals with the identification of various details about the author of the text (e.g., age and gender). In this paper, we describe the participation of our team (hacohenkerner19) in the PAN 2019 shared task on Bots and Gender Profiling in two languages: English and Spanish. Given a Twitter feed, we should determine whether its author is a bot or a human. In the case of human, we should identify her/his gender. In this paper, we describe our preprocessing methods, feature sets, five applied machine learning methods, and accuracy results. The best accuracy result for the English dataset (84.8%) was obtained by LinearSVC using 2,000 word unigrams. The same result (84.8%) was also obtained by LR by using four preprocessing methods, 2,000 word unigrams, and 1,000 word bigrams with maximal skips of 2 words. The best accuracy result (75,54%) for the Spanish dataset was achieved using LinearSVC with only the HTML tag removal preprocessing method and a combination of 1,000 word unigrams, 1,000 word bigrams, and 3,000 character trigrams.

AB - Author profiling deals with the identification of various details about the author of the text (e.g., age and gender). In this paper, we describe the participation of our team (hacohenkerner19) in the PAN 2019 shared task on Bots and Gender Profiling in two languages: English and Spanish. Given a Twitter feed, we should determine whether its author is a bot or a human. In the case of human, we should identify her/his gender. In this paper, we describe our preprocessing methods, feature sets, five applied machine learning methods, and accuracy results. The best accuracy result for the English dataset (84.8%) was obtained by LinearSVC using 2,000 word unigrams. The same result (84.8%) was also obtained by LR by using four preprocessing methods, 2,000 word unigrams, and 1,000 word bigrams with maximal skips of 2 words. The best accuracy result (75,54%) for the Spanish dataset was achieved using LinearSVC with only the HTML tag removal preprocessing method and a combination of 1,000 word unigrams, 1,000 word bigrams, and 3,000 character trigrams.

KW - Bot Profiling

KW - Character N-grams

KW - Gender Profiling

KW - Supervised Machine Learning

KW - Word N-grams

UR - https://www.scopus.com/pages/publications/85070489908

M3 - ???researchoutput.researchoutputtypes.contributiontojournal.conferencearticle???

AN - SCOPUS:85070489908

SN - 1613-0073

VL - 2380

JO - CEUR Workshop Proceedings

JF - CEUR Workshop Proceedings

T2 - 20th Working Notes of CLEF Conference and Labs of the Evaluation Forum, CLEF 2019

Y2 - 9 September 2019 through 12 September 2019

ER -

Bots and gender profiling of tweets using word and character N-grams notebook for PAN at CLEF 2019

Abstract

Bibliographical note

Funding

Keywords

Other files and links

Fingerprint

Cite this