DCWEB-SOBA: Deep Contextual Word Embeddings-Based Semi-automatic Ontology Building for Aspect-Based Sentiment Classification

Roos van Lookeren Campagne; David van Ommen; Mark Rademaker; Tom Teurlings; Flavius Frasincar

doi:10.1007/978-3-031-06981-9_11

DCWEB-SOBA: Deep Contextual Word Embeddings-Based Semi-automatic Ontology Building for Aspect-Based Sentiment Classification

Roos van Lookeren Campagne
, David van Ommen
, Mark Rademaker
, Tom Teurlings
, Flavius Frasincar^*

^*Corresponding author for this work

Econometrics

Erasmus School of Health Policy and Management (ESHPM)

Research output: Chapter/Conference proceeding › Conference proceeding › Academic › peer-review

2 Citations (Scopus)

Abstract

In this paper, we propose the use of deep contextualised word embeddings to semi-automatically build a domain sentiment ontology. Compared to previous research, we use deep contextualised word embeddings to better cope with various meanings of words. A state-of-the-art hybrid method is used for aspect-based sentiment analysis, called HAABSA++, to evaluate our obtained ontology on the SemEval-2016 restaurant dataset. We achieve a prediction accuracy of 81.85% for the hybrid model with our ontology, which outperforms the hybrid model with other considered ontologies. Furthermore, we find that the ontology obtained from our proposed domain sentiment ontology builder, called DCWEB-SOBA, on itself improves the accuracy for the conclusive cases from 83.04% to 84.52% compared to the ontology builder based on non-contextual word embeddings, WEB-SOBA.

Original language	English
Title of host publication	The Semantic Web - 19th International Conference, ESWC 2022, Proceedings
Editors	Paul Groth, Maria-Esther Vidal, Fabian Suchanek, Pedro Szekley, Pavan Kapanipathi, Catia Pesquita, Hala Skaf-Molli, Minna Tamper
Publisher	Springer Science+Business Media
Pages	183-199
Number of pages	17
Volume	13261
ISBN (Print)	9783031069802
DOIs	https://doi.org/10.1007/978-3-031-06981-9_11
Publication status	Published - 2022
Event	19th International Conference on European Semantic Web Conference, ESWC 2022 - Hersonissos, Greece Duration: 29 May 2022 → 2 Jun 2022

Publication series

Series	Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume	13261 LNCS
ISSN	0302-9743

Conference

Conference	19th International Conference on European Semantic Web Conference, ESWC 2022
Country/Territory	Greece
City	Hersonissos
Period	29/05/22 → 2/06/22

Bibliographical note

Publisher Copyright:
© 2022, The Author(s), under exclusive license to Springer Nature Switzerland AG.

Access to Document

10.1007/978-3-031-06981-9_11

Cite this

van Lookeren Campagne, R., van Ommen, D., Rademaker, M., Teurlings, T., & Frasincar, F. (2022). DCWEB-SOBA: Deep Contextual Word Embeddings-Based Semi-automatic Ontology Building for Aspect-Based Sentiment Classification. In P. Groth, M.-E. Vidal, F. Suchanek, P. Szekley, P. Kapanipathi, C. Pesquita, H. Skaf-Molli, & M. Tamper (Eds.), The Semantic Web - 19th International Conference, ESWC 2022, Proceedings (Vol. 13261, pp. 183-199). Springer Science+Business Media. https://doi.org/10.1007/978-3-031-06981-9_11

van Lookeren Campagne, Roos ; van Ommen, David ; Rademaker, Mark et al. / DCWEB-SOBA : Deep Contextual Word Embeddings-Based Semi-automatic Ontology Building for Aspect-Based Sentiment Classification. The Semantic Web - 19th International Conference, ESWC 2022, Proceedings. editor / Paul Groth ; Maria-Esther Vidal ; Fabian Suchanek ; Pedro Szekley ; Pavan Kapanipathi ; Catia Pesquita ; Hala Skaf-Molli ; Minna Tamper. Vol. 13261 Springer Science+Business Media, 2022. pp. 183-199 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), Vol. 13261 LNCS).

@inproceedings{5b5451b2a0d1400a8914a1c6ab6ebb92,

title = "DCWEB-SOBA: Deep Contextual Word Embeddings-Based Semi-automatic Ontology Building for Aspect-Based Sentiment Classification",

abstract = "In this paper, we propose the use of deep contextualised word embeddings to semi-automatically build a domain sentiment ontology. Compared to previous research, we use deep contextualised word embeddings to better cope with various meanings of words. A state-of-the-art hybrid method is used for aspect-based sentiment analysis, called HAABSA++, to evaluate our obtained ontology on the SemEval-2016 restaurant dataset. We achieve a prediction accuracy of 81.85\% for the hybrid model with our ontology, which outperforms the hybrid model with other considered ontologies. Furthermore, we find that the ontology obtained from our proposed domain sentiment ontology builder, called DCWEB-SOBA, on itself improves the accuracy for the conclusive cases from 83.04\% to 84.52\% compared to the ontology builder based on non-contextual word embeddings, WEB-SOBA.",

author = "\{van Lookeren Campagne\}, Roos and \{van Ommen\}, David and Mark Rademaker and Tom Teurlings and Flavius Frasincar",

note = "Publisher Copyright: {\textcopyright} 2022, The Author(s), under exclusive license to Springer Nature Switzerland AG.; 19th International Conference on European Semantic Web Conference, ESWC 2022 ; Conference date: 29-05-2022 Through 02-06-2022",

year = "2022",

doi = "10.1007/978-3-031-06981-9\_11",

language = "English",

isbn = "9783031069802",

volume = "13261",

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

publisher = "Springer Science+Business Media",

pages = "183--199",

editor = "Paul Groth and Maria-Esther Vidal and Fabian Suchanek and Pedro Szekley and Pavan Kapanipathi and Catia Pesquita and Hala Skaf-Molli and Minna Tamper",

booktitle = "The Semantic Web - 19th International Conference, ESWC 2022, Proceedings",

address = "Germany",

}

van Lookeren Campagne, R, van Ommen, D, Rademaker, M, Teurlings, T & Frasincar, F 2022, DCWEB-SOBA: Deep Contextual Word Embeddings-Based Semi-automatic Ontology Building for Aspect-Based Sentiment Classification. in P Groth, M-E Vidal, F Suchanek, P Szekley, P Kapanipathi, C Pesquita, H Skaf-Molli & M Tamper (eds), The Semantic Web - 19th International Conference, ESWC 2022, Proceedings. vol. 13261, Springer Science+Business Media, Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 13261 LNCS, pp. 183-199, 19th International Conference on European Semantic Web Conference, ESWC 2022, Hersonissos, Greece, 29/05/22. https://doi.org/10.1007/978-3-031-06981-9_11

DCWEB-SOBA: Deep Contextual Word Embeddings-Based Semi-automatic Ontology Building for Aspect-Based Sentiment Classification. / van Lookeren Campagne, Roos; van Ommen, David; Rademaker, Mark et al.
The Semantic Web - 19th International Conference, ESWC 2022, Proceedings. ed. / Paul Groth; Maria-Esther Vidal; Fabian Suchanek; Pedro Szekley; Pavan Kapanipathi; Catia Pesquita; Hala Skaf-Molli; Minna Tamper. Vol. 13261 Springer Science+Business Media, 2022. p. 183-199 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), Vol. 13261 LNCS).

Research output: Chapter/Conference proceeding › Conference proceeding › Academic › peer-review

TY - GEN

T1 - DCWEB-SOBA

T2 - 19th International Conference on European Semantic Web Conference, ESWC 2022

AU - van Lookeren Campagne, Roos

AU - van Ommen, David

AU - Rademaker, Mark

AU - Teurlings, Tom

AU - Frasincar, Flavius

PY - 2022

Y1 - 2022

N2 - In this paper, we propose the use of deep contextualised word embeddings to semi-automatically build a domain sentiment ontology. Compared to previous research, we use deep contextualised word embeddings to better cope with various meanings of words. A state-of-the-art hybrid method is used for aspect-based sentiment analysis, called HAABSA++, to evaluate our obtained ontology on the SemEval-2016 restaurant dataset. We achieve a prediction accuracy of 81.85% for the hybrid model with our ontology, which outperforms the hybrid model with other considered ontologies. Furthermore, we find that the ontology obtained from our proposed domain sentiment ontology builder, called DCWEB-SOBA, on itself improves the accuracy for the conclusive cases from 83.04% to 84.52% compared to the ontology builder based on non-contextual word embeddings, WEB-SOBA.

AB - In this paper, we propose the use of deep contextualised word embeddings to semi-automatically build a domain sentiment ontology. Compared to previous research, we use deep contextualised word embeddings to better cope with various meanings of words. A state-of-the-art hybrid method is used for aspect-based sentiment analysis, called HAABSA++, to evaluate our obtained ontology on the SemEval-2016 restaurant dataset. We achieve a prediction accuracy of 81.85% for the hybrid model with our ontology, which outperforms the hybrid model with other considered ontologies. Furthermore, we find that the ontology obtained from our proposed domain sentiment ontology builder, called DCWEB-SOBA, on itself improves the accuracy for the conclusive cases from 83.04% to 84.52% compared to the ontology builder based on non-contextual word embeddings, WEB-SOBA.

UR - https://www.scopus.com/pages/publications/85131928135

U2 - 10.1007/978-3-031-06981-9_11

DO - 10.1007/978-3-031-06981-9_11

M3 - Conference proceeding

AN - SCOPUS:85131928135

SN - 9783031069802

VL - 13261

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

SP - 183

EP - 199

BT - The Semantic Web - 19th International Conference, ESWC 2022, Proceedings

A2 - Groth, Paul

A2 - Vidal, Maria-Esther

A2 - Suchanek, Fabian

A2 - Szekley, Pedro

A2 - Kapanipathi, Pavan

A2 - Pesquita, Catia

A2 - Skaf-Molli, Hala

A2 - Tamper, Minna

PB - Springer Science+Business Media

Y2 - 29 May 2022 through 2 June 2022

ER -

van Lookeren Campagne R, van Ommen D, Rademaker M, Teurlings T, Frasincar F. DCWEB-SOBA: Deep Contextual Word Embeddings-Based Semi-automatic Ontology Building for Aspect-Based Sentiment Classification. In Groth P, Vidal ME, Suchanek F, Szekley P, Kapanipathi P, Pesquita C, Skaf-Molli H, Tamper M, editors, The Semantic Web - 19th International Conference, ESWC 2022, Proceedings. Vol. 13261. Springer Science+Business Media. 2022. p. 183-199. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), Vol. 13261 LNCS). doi: 10.1007/978-3-031-06981-9_11