Adversarial Training for a Hybrid Approach to Aspect-Based Sentiment Analysis

Ron Hochstenbach; Flavius Frasincar; Maria Mihaela Truşcǎ

doi:10.1007/978-3-030-91560-5_21

Adversarial Training for a Hybrid Approach to Aspect-Based Sentiment Analysis

Ron Hochstenbach, Flavius Frasincar, Maria Mihaela Truşcǎ^*

^*Corresponding author for this work

Econometrics

Research output: Chapter/Conference proceeding › Chapter › Academic

Abstract

The increasing popularity of the Web has subsequently increased the abundance of reviews on products and services. Mining these reviews for expressed sentiment is beneficial for both companies and consumers, as quality can be improved based on this information. In this paper, we consider the state-of-the-art HAABSA++ algorithm for aspect-based sentiment analysis tasked with identifying the sentiment expressed towards a given aspect in review sentences. Specifically, we train the neural network part of this algorithm using an adversarial network, a novel machine learning training method where a generator network tries to fool the classifier network by generating highly realistic new samples, as such increasing robustness. This method, as of yet never in its classical form applied to aspect-based sentiment analysis, is found to be able to considerably improve the out-of-sample accuracy of HAABSA++: for the SemEval 2015 dataset, accuracy was increased from 81.7% to 82.5%, and for the SemEval 2016 task, accuracy increased from 84.4% to 87.3%.

Original language	English
Title of host publication	22nd International Conference on Web Information Systems Engineering (WISE 2021)
Editors	Wenjie Zhang, Lei Zou, Zakaria Maamar, Lu Chen
Publisher	Springer-Verlag
Pages	291-305
Number of pages	15
ISBN (Print)	9783030915599
DOIs	https://doi.org/10.1007/978-3-030-91560-5_21
Publication status	Published - 2021
Event	22nd International Conference on Web Information Systems Engineering, WISE 2021 - Melbourne, Australia Duration: 26 Oct 2021 → 29 Oct 2021

Publication series

Series	Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume	13081 LNCS
ISSN	0302-9743

Conference

Conference	22nd International Conference on Web Information Systems Engineering, WISE 2021
Country/Territory	Australia
City	Melbourne
Period	26/10/21 → 29/10/21

Bibliographical note

Publisher Copyright:
© 2021, Springer Nature Switzerland AG.

Research programs

ESE - E&MS

Access to Document

10.1007/978-3-030-91560-5_21

Cite this

Hochstenbach, Ron ; Frasincar, Flavius ; Truşcǎ, Maria Mihaela. / Adversarial Training for a Hybrid Approach to Aspect-Based Sentiment Analysis. 22nd International Conference on Web Information Systems Engineering (WISE 2021). editor / Wenjie Zhang ; Lei Zou ; Zakaria Maamar ; Lu Chen. Springer-Verlag, 2021. pp. 291-305 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), Vol. 13081 LNCS).

@inbook{852d120aced0438ca1719dee05dbb79e,

title = "Adversarial Training for a Hybrid Approach to Aspect-Based Sentiment Analysis",

abstract = "The increasing popularity of the Web has subsequently increased the abundance of reviews on products and services. Mining these reviews for expressed sentiment is beneficial for both companies and consumers, as quality can be improved based on this information. In this paper, we consider the state-of-the-art HAABSA++ algorithm for aspect-based sentiment analysis tasked with identifying the sentiment expressed towards a given aspect in review sentences. Specifically, we train the neural network part of this algorithm using an adversarial network, a novel machine learning training method where a generator network tries to fool the classifier network by generating highly realistic new samples, as such increasing robustness. This method, as of yet never in its classical form applied to aspect-based sentiment analysis, is found to be able to considerably improve the out-of-sample accuracy of HAABSA++: for the SemEval 2015 dataset, accuracy was increased from 81.7% to 82.5%, and for the SemEval 2016 task, accuracy increased from 84.4% to 87.3%.",

author = "Ron Hochstenbach and Flavius Frasincar and Tru{\c s}cǎ, {Maria Mihaela}",

note = "Publisher Copyright: {\textcopyright} 2021, Springer Nature Switzerland AG.; 22nd International Conference on Web Information Systems Engineering, WISE 2021 ; Conference date: 26-10-2021 Through 29-10-2021",

year = "2021",

doi = "10.1007/978-3-030-91560-5_21",

language = "English",

isbn = "9783030915599",

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

publisher = "Springer-Verlag",

pages = "291--305",

editor = "Wenjie Zhang and Lei Zou and Zakaria Maamar and Lu Chen",

booktitle = "22nd International Conference on Web Information Systems Engineering (WISE 2021)",

address = "Germany",

}

Hochstenbach, R, Frasincar, F & Truşcǎ, MM 2021, Adversarial Training for a Hybrid Approach to Aspect-Based Sentiment Analysis. in W Zhang, L Zou, Z Maamar & L Chen (eds), 22nd International Conference on Web Information Systems Engineering (WISE 2021). Springer-Verlag, Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 13081 LNCS, pp. 291-305, 22nd International Conference on Web Information Systems Engineering, WISE 2021, Melbourne, Australia, 26/10/21. https://doi.org/10.1007/978-3-030-91560-5_21

Adversarial Training for a Hybrid Approach to Aspect-Based Sentiment Analysis. / Hochstenbach, Ron; Frasincar, Flavius; Truşcǎ, Maria Mihaela.
22nd International Conference on Web Information Systems Engineering (WISE 2021). ed. / Wenjie Zhang; Lei Zou; Zakaria Maamar; Lu Chen. Springer-Verlag, 2021. p. 291-305 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), Vol. 13081 LNCS).

Research output: Chapter/Conference proceeding › Chapter › Academic

TY - CHAP

T1 - Adversarial Training for a Hybrid Approach to Aspect-Based Sentiment Analysis

AU - Hochstenbach, Ron

AU - Frasincar, Flavius

AU - Truşcǎ, Maria Mihaela

PY - 2021

Y1 - 2021

N2 - The increasing popularity of the Web has subsequently increased the abundance of reviews on products and services. Mining these reviews for expressed sentiment is beneficial for both companies and consumers, as quality can be improved based on this information. In this paper, we consider the state-of-the-art HAABSA++ algorithm for aspect-based sentiment analysis tasked with identifying the sentiment expressed towards a given aspect in review sentences. Specifically, we train the neural network part of this algorithm using an adversarial network, a novel machine learning training method where a generator network tries to fool the classifier network by generating highly realistic new samples, as such increasing robustness. This method, as of yet never in its classical form applied to aspect-based sentiment analysis, is found to be able to considerably improve the out-of-sample accuracy of HAABSA++: for the SemEval 2015 dataset, accuracy was increased from 81.7% to 82.5%, and for the SemEval 2016 task, accuracy increased from 84.4% to 87.3%.

AB - The increasing popularity of the Web has subsequently increased the abundance of reviews on products and services. Mining these reviews for expressed sentiment is beneficial for both companies and consumers, as quality can be improved based on this information. In this paper, we consider the state-of-the-art HAABSA++ algorithm for aspect-based sentiment analysis tasked with identifying the sentiment expressed towards a given aspect in review sentences. Specifically, we train the neural network part of this algorithm using an adversarial network, a novel machine learning training method where a generator network tries to fool the classifier network by generating highly realistic new samples, as such increasing robustness. This method, as of yet never in its classical form applied to aspect-based sentiment analysis, is found to be able to considerably improve the out-of-sample accuracy of HAABSA++: for the SemEval 2015 dataset, accuracy was increased from 81.7% to 82.5%, and for the SemEval 2016 task, accuracy increased from 84.4% to 87.3%.

UR - http://www.scopus.com/inward/record.url?scp=85121930542&partnerID=8YFLogxK

U2 - 10.1007/978-3-030-91560-5_21

DO - 10.1007/978-3-030-91560-5_21

M3 - Chapter

AN - SCOPUS:85121930542

SN - 9783030915599

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

SP - 291

EP - 305

BT - 22nd International Conference on Web Information Systems Engineering (WISE 2021)

A2 - Zhang, Wenjie

A2 - Zou, Lei

A2 - Maamar, Zakaria

A2 - Chen, Lu

PB - Springer-Verlag

T2 - 22nd International Conference on Web Information Systems Engineering, WISE 2021

Y2 - 26 October 2021 through 29 October 2021

ER -

Hochstenbach R, Frasincar F, Truşcǎ MM. Adversarial Training for a Hybrid Approach to Aspect-Based Sentiment Analysis. In Zhang W, Zou L, Maamar Z, Chen L, editors, 22nd International Conference on Web Information Systems Engineering (WISE 2021). Springer-Verlag. 2021. p. 291-305. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), Vol. 13081 LNCS). doi: 10.1007/978-3-030-91560-5_21