Knowledge Injection for Aspect-Based Sentiment Classification

Romany Dekker; Danae Gielisse; Chaya Jaggan; Sander Meijers; Flavius Frasincar

doi:10.1007/978-3-031-39821-6_14

Knowledge Injection for Aspect-Based Sentiment Classification

Romany Dekker
, Danae Gielisse
, Chaya Jaggan
, Sander Meijers
, Flavius Frasincar^*

^*Corresponding author for this work

Econometrics

Erasmus University Rotterdam

Research output: Chapter/Conference proceeding › Chapter › Academic

3 Citations (Scopus)

Abstract

Since the increase of Web reviews of products and services, Aspect-Based Sentiment Classification (ABSC) has become more important to determine the sentiment of online opinions. Useful information extracted from these reviews can then be used by companies themselves, but can also be applicable by consumers. In the recent literature on ABSC, hybrid methods, which combine knowledge-based and machine learning approaches, are becoming more popular as well. However, in this work, instead of following a two-step procedure, we attempt to improve the model accuracy by proposing to directly inject the information from a domain ontology in a state-of-the-art neural network model, more precisely LCR-Rot-hop++. Furthermore, by using soft-positioning and visible matrices we aim to prevent that the injected knowledge hinders the semantics of the original sentences. To evaluate the accuracy of our model, LCR-Rot-hop-ont++, we use the standard SemEval 2015 and SemEval 2016 datasets for ABSC. We conclude that knowledge injection in the neural network is effective for sentiment classification, especially if the amount of labeled data is limited.

Original language	English
Title of host publication	Database and Expert Systems Applications - 34th International Conference, DEXA 2023, Proceedings
Editors	Christine Strauss, Toshiyuki Amagasa, Gabriele Kotsis, Ismail Khalil, A Min Tjoa
Publisher	Springer Science+Business Media
Pages	173-187
Number of pages	15
ISBN (Print)	9783031398209
DOIs	https://doi.org/10.1007/978-3-031-39821-6_14
Publication status	Published - 2023
Event	The 34th International Conference on Database and Expert Systems Applications DEXA 2023 - Penang, Malaysia Duration: 28 Aug 2023 → 30 Aug 2023

Publication series

Series	Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume	14147 LNCS
ISSN	0302-9743

Conference

Conference	The 34th International Conference on Database and Expert Systems Applications DEXA 2023
Country/Territory	Malaysia
City	Penang
Period	28/08/23 → 30/08/23

Bibliographical note

Publisher Copyright:
© 2023, The Author(s), under exclusive license to Springer Nature Switzerland AG.

Access to Document

10.1007/978-3-031-39821-6_14

Cite this

Dekker, R., Gielisse, D., Jaggan, C., Meijers, S., & Frasincar, F. (2023). Knowledge Injection for Aspect-Based Sentiment Classification. In C. Strauss, T. Amagasa, G. Kotsis, I. Khalil, & A. M. Tjoa (Eds.), Database and Expert Systems Applications - 34th International Conference, DEXA 2023, Proceedings (pp. 173-187). Springer Science+Business Media. https://doi.org/10.1007/978-3-031-39821-6_14

Dekker, Romany ; Gielisse, Danae ; Jaggan, Chaya et al. / Knowledge Injection for Aspect-Based Sentiment Classification. Database and Expert Systems Applications - 34th International Conference, DEXA 2023, Proceedings. editor / Christine Strauss ; Toshiyuki Amagasa ; Gabriele Kotsis ; Ismail Khalil ; A Min Tjoa. Springer Science+Business Media, 2023. pp. 173-187 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), Vol. 14147 LNCS).

@inbook{c0b08c5f8ca849d69f5b6b9f4b3676c6,

title = "Knowledge Injection for Aspect-Based Sentiment Classification",

abstract = "Since the increase of Web reviews of products and services, Aspect-Based Sentiment Classification (ABSC) has become more important to determine the sentiment of online opinions. Useful information extracted from these reviews can then be used by companies themselves, but can also be applicable by consumers. In the recent literature on ABSC, hybrid methods, which combine knowledge-based and machine learning approaches, are becoming more popular as well. However, in this work, instead of following a two-step procedure, we attempt to improve the model accuracy by proposing to directly inject the information from a domain ontology in a state-of-the-art neural network model, more precisely LCR-Rot-hop++. Furthermore, by using soft-positioning and visible matrices we aim to prevent that the injected knowledge hinders the semantics of the original sentences. To evaluate the accuracy of our model, LCR-Rot-hop-ont++, we use the standard SemEval 2015 and SemEval 2016 datasets for ABSC. We conclude that knowledge injection in the neural network is effective for sentiment classification, especially if the amount of labeled data is limited.",

author = "Romany Dekker and Danae Gielisse and Chaya Jaggan and Sander Meijers and Flavius Frasincar",

note = "Publisher Copyright: {\textcopyright} 2023, The Author(s), under exclusive license to Springer Nature Switzerland AG.; The 34th International Conference on Database and Expert Systems Applications DEXA 2023 ; Conference date: 28-08-2023 Through 30-08-2023",

year = "2023",

doi = "10.1007/978-3-031-39821-6\_14",

language = "English",

isbn = "9783031398209",

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

publisher = "Springer Science+Business Media",

pages = "173--187",

editor = "Christine Strauss and Toshiyuki Amagasa and Gabriele Kotsis and Ismail Khalil and Tjoa, \{A Min\}",

booktitle = "Database and Expert Systems Applications - 34th International Conference, DEXA 2023, Proceedings",

address = "Germany",

}

Dekker, R, Gielisse, D, Jaggan, C, Meijers, S & Frasincar, F 2023, Knowledge Injection for Aspect-Based Sentiment Classification. in C Strauss, T Amagasa, G Kotsis, I Khalil & AM Tjoa (eds), Database and Expert Systems Applications - 34th International Conference, DEXA 2023, Proceedings. Springer Science+Business Media, Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 14147 LNCS, pp. 173-187, The 34th International Conference on Database and Expert Systems Applications DEXA 2023, Penang, Malaysia, 28/08/23. https://doi.org/10.1007/978-3-031-39821-6_14

Knowledge Injection for Aspect-Based Sentiment Classification. / Dekker, Romany; Gielisse, Danae; Jaggan, Chaya et al.
Database and Expert Systems Applications - 34th International Conference, DEXA 2023, Proceedings. ed. / Christine Strauss; Toshiyuki Amagasa; Gabriele Kotsis; Ismail Khalil; A Min Tjoa. Springer Science+Business Media, 2023. p. 173-187 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), Vol. 14147 LNCS).

Research output: Chapter/Conference proceeding › Chapter › Academic

TY - CHAP

T1 - Knowledge Injection for Aspect-Based Sentiment Classification

AU - Dekker, Romany

AU - Gielisse, Danae

AU - Jaggan, Chaya

AU - Meijers, Sander

AU - Frasincar, Flavius

PY - 2023

Y1 - 2023

N2 - Since the increase of Web reviews of products and services, Aspect-Based Sentiment Classification (ABSC) has become more important to determine the sentiment of online opinions. Useful information extracted from these reviews can then be used by companies themselves, but can also be applicable by consumers. In the recent literature on ABSC, hybrid methods, which combine knowledge-based and machine learning approaches, are becoming more popular as well. However, in this work, instead of following a two-step procedure, we attempt to improve the model accuracy by proposing to directly inject the information from a domain ontology in a state-of-the-art neural network model, more precisely LCR-Rot-hop++. Furthermore, by using soft-positioning and visible matrices we aim to prevent that the injected knowledge hinders the semantics of the original sentences. To evaluate the accuracy of our model, LCR-Rot-hop-ont++, we use the standard SemEval 2015 and SemEval 2016 datasets for ABSC. We conclude that knowledge injection in the neural network is effective for sentiment classification, especially if the amount of labeled data is limited.

AB - Since the increase of Web reviews of products and services, Aspect-Based Sentiment Classification (ABSC) has become more important to determine the sentiment of online opinions. Useful information extracted from these reviews can then be used by companies themselves, but can also be applicable by consumers. In the recent literature on ABSC, hybrid methods, which combine knowledge-based and machine learning approaches, are becoming more popular as well. However, in this work, instead of following a two-step procedure, we attempt to improve the model accuracy by proposing to directly inject the information from a domain ontology in a state-of-the-art neural network model, more precisely LCR-Rot-hop++. Furthermore, by using soft-positioning and visible matrices we aim to prevent that the injected knowledge hinders the semantics of the original sentences. To evaluate the accuracy of our model, LCR-Rot-hop-ont++, we use the standard SemEval 2015 and SemEval 2016 datasets for ABSC. We conclude that knowledge injection in the neural network is effective for sentiment classification, especially if the amount of labeled data is limited.

UR - https://www.scopus.com/pages/publications/85174682704

U2 - 10.1007/978-3-031-39821-6_14

DO - 10.1007/978-3-031-39821-6_14

M3 - Chapter

AN - SCOPUS:85174682704

SN - 9783031398209

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

SP - 173

EP - 187

BT - Database and Expert Systems Applications - 34th International Conference, DEXA 2023, Proceedings

A2 - Strauss, Christine

A2 - Amagasa, Toshiyuki

A2 - Kotsis, Gabriele

A2 - Khalil, Ismail

A2 - Tjoa, A Min

PB - Springer Science+Business Media

T2 - The 34th International Conference on Database and Expert Systems Applications DEXA 2023

Y2 - 28 August 2023 through 30 August 2023

ER -

Dekker R, Gielisse D, Jaggan C, Meijers S, Frasincar F. Knowledge Injection for Aspect-Based Sentiment Classification. In Strauss C, Amagasa T, Kotsis G, Khalil I, Tjoa AM, editors, Database and Expert Systems Applications - 34th International Conference, DEXA 2023, Proceedings. Springer Science+Business Media. 2023. p. 173-187. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), Vol. 14147 LNCS). doi: 10.1007/978-3-031-39821-6_14