Detecting speculations, contrasts and conditionals in consumer reviews

Maria Skeppstedt; Teri Schamp-Bjerede; Magnus Sahlgren; Carita Paradis; Andreas Kerren

Detecting speculations, contrasts and conditionals in consumer reviews

Maria Skeppstedt, Teri Schamp-Bjerede, Magnus Sahlgren, Carita Paradis, Andreas Kerren

English Studies

Gagavai AB

Research output: Chapter in Book/Report/Conference proceeding › Paper in conference proceeding › peer-review

Abstract

A support vector classifier was compared to a lexicon-based approach for the task of detecting the stance categories speculation, contrast and conditional in English consumer reviews. Around 3,000 training instances were required to achieve a stable performance of an F-score of 90 for speculation. This outperformed the lexicon-based approach, for which an Fscore of just above 80 was achieved. The machine learning results for the other two categories showed a lower average (an approximate F-score of 60 for contrast and 70 for conditional), as well as a larger variance, and were only slightly better than lexicon matching. Therefore, while machine learning was successful for detecting speculation, a well-curated lexicon might be a more suitable approach for detecting contrast and conditional.

Original language	English
Title of host publication	6th Workshop on Computational Approaches to Subjectivity, Sentiment & Social Media Analysis, WASSA 2015 : Workshop proceedings
Editors	Balahur Alexandra, van der Goot Erik, Vossen Piek, Montoyo Andrés
Publisher	Association for Computational Linguistics
Pages	162-168
Number of pages	7
ISBN (Print)	978-1-941643-32-7
Publication status	Published - 2015
Event	6th Workshop on Computational Approaches to Subjectivity, Sentiment & Social Media Analysis (WASSA '15) - Lisbon Duration: 2015 Sept 17 → …

Conference

Conference	6th Workshop on Computational Approaches to Subjectivity, Sentiment & Social Media Analysis (WASSA '15)
Period	2015/09/17 → …

Subject classification (UKÄ)

Languages and Literature
Computer Science

Free keywords

consumer reviews
support vector classifier
stance

Access to Document

http://www.aclweb.org/anthology/W/W15/W15-29.pdf#page.162

StaViCTA - Advances in the description and explanation of stance in discourse using visual and computational text analytics
Paradis, C. (PI), Kerren, A. (PI), Sahlgren, M. (PI), Kucher, K. (Researcher), Skeppstedt, M. (Researcher) & Simaki, V. (PI)
Swedish Research Council
2013/01/01 → 2018/01/02
Project: Research

Cite this

Skeppstedt, M., Schamp-Bjerede, T., Sahlgren, M., Paradis, C., & Kerren, A. (2015). Detecting speculations, contrasts and conditionals in consumer reviews. In B. Alexandra, V. D. G. Erik, V. Piek, & M. Andrés (Eds.), 6th Workshop on Computational Approaches to Subjectivity, Sentiment & Social Media Analysis, WASSA 2015 : Workshop proceedings (pp. 162-168). Association for Computational Linguistics. http://www.aclweb.org/anthology/W/W15/W15-29.pdf#page.162

Skeppstedt, Maria ; Schamp-Bjerede, Teri ; Sahlgren, Magnus et al. / Detecting speculations, contrasts and conditionals in consumer reviews. 6th Workshop on Computational Approaches to Subjectivity, Sentiment & Social Media Analysis, WASSA 2015 : Workshop proceedings. editor / Balahur Alexandra ; van der Goot Erik ; Vossen Piek ; Montoyo Andrés. Association for Computational Linguistics, 2015. pp. 162-168

@inproceedings{900a4c9b8c5348c086797aef36931455,

title = "Detecting speculations, contrasts and conditionals in consumer reviews",

abstract = "A support vector classifier was compared to a lexicon-based approach for the task of detecting the stance categories speculation, contrast and conditional in English consumer reviews. Around 3,000 training instances were required to achieve a stable performance of an F-score of 90 for speculation. This outperformed the lexicon-based approach, for which an Fscore of just above 80 was achieved. The machine learning results for the other two categories showed a lower average (an approximate F-score of 60 for contrast and 70 for conditional), as well as a larger variance, and were only slightly better than lexicon matching. Therefore, while machine learning was successful for detecting speculation, a well-curated lexicon might be a more suitable approach for detecting contrast and conditional.",

keywords = "consumer reviews, support vector classifier, stance",

author = "Maria Skeppstedt and Teri Schamp-Bjerede and Magnus Sahlgren and Carita Paradis and Andreas Kerren",

year = "2015",

language = "English",

isbn = "978-1-941643-32-7",

pages = "162--168",

editor = "Balahur Alexandra and Erik, {van der Goot} and Vossen Piek and Montoyo Andr{\'e}s",

booktitle = "6th Workshop on Computational Approaches to Subjectivity, Sentiment & Social Media Analysis, WASSA 2015 : Workshop proceedings",

publisher = "Association for Computational Linguistics",

address = "United States",

note = "6th Workshop on Computational Approaches to Subjectivity, Sentiment & Social Media Analysis (WASSA '15) ; Conference date: 17-09-2015",

}

Skeppstedt, M, Schamp-Bjerede, T, Sahlgren, M, Paradis, C & Kerren, A 2015, Detecting speculations, contrasts and conditionals in consumer reviews. in B Alexandra, VDG Erik, V Piek & M Andrés (eds), 6th Workshop on Computational Approaches to Subjectivity, Sentiment & Social Media Analysis, WASSA 2015 : Workshop proceedings. Association for Computational Linguistics, pp. 162-168, 6th Workshop on Computational Approaches to Subjectivity, Sentiment & Social Media Analysis (WASSA '15), 2015/09/17. <http://www.aclweb.org/anthology/W/W15/W15-29.pdf#page.162>

Detecting speculations, contrasts and conditionals in consumer reviews. / Skeppstedt, Maria; Schamp-Bjerede, Teri; Sahlgren, Magnus et al.
6th Workshop on Computational Approaches to Subjectivity, Sentiment & Social Media Analysis, WASSA 2015 : Workshop proceedings. ed. / Balahur Alexandra; van der Goot Erik; Vossen Piek; Montoyo Andrés. Association for Computational Linguistics, 2015. p. 162-168.

Research output: Chapter in Book/Report/Conference proceeding › Paper in conference proceeding › peer-review

TY - GEN

T1 - Detecting speculations, contrasts and conditionals in consumer reviews

AU - Skeppstedt, Maria

AU - Schamp-Bjerede, Teri

AU - Sahlgren, Magnus

AU - Paradis, Carita

AU - Kerren, Andreas

PY - 2015

Y1 - 2015

N2 - A support vector classifier was compared to a lexicon-based approach for the task of detecting the stance categories speculation, contrast and conditional in English consumer reviews. Around 3,000 training instances were required to achieve a stable performance of an F-score of 90 for speculation. This outperformed the lexicon-based approach, for which an Fscore of just above 80 was achieved. The machine learning results for the other two categories showed a lower average (an approximate F-score of 60 for contrast and 70 for conditional), as well as a larger variance, and were only slightly better than lexicon matching. Therefore, while machine learning was successful for detecting speculation, a well-curated lexicon might be a more suitable approach for detecting contrast and conditional.

AB - A support vector classifier was compared to a lexicon-based approach for the task of detecting the stance categories speculation, contrast and conditional in English consumer reviews. Around 3,000 training instances were required to achieve a stable performance of an F-score of 90 for speculation. This outperformed the lexicon-based approach, for which an Fscore of just above 80 was achieved. The machine learning results for the other two categories showed a lower average (an approximate F-score of 60 for contrast and 70 for conditional), as well as a larger variance, and were only slightly better than lexicon matching. Therefore, while machine learning was successful for detecting speculation, a well-curated lexicon might be a more suitable approach for detecting contrast and conditional.

KW - consumer reviews

KW - support vector classifier

KW - stance

M3 - Paper in conference proceeding

SN - 978-1-941643-32-7

SP - 162

EP - 168

BT - 6th Workshop on Computational Approaches to Subjectivity, Sentiment & Social Media Analysis, WASSA 2015 : Workshop proceedings

A2 - Alexandra, Balahur

A2 - Erik, van der Goot

A2 - Piek, Vossen

A2 - Andrés, Montoyo

PB - Association for Computational Linguistics

T2 - 6th Workshop on Computational Approaches to Subjectivity, Sentiment & Social Media Analysis (WASSA '15)

Y2 - 17 September 2015

ER -

Skeppstedt M, Schamp-Bjerede T, Sahlgren M, Paradis C, Kerren A. Detecting speculations, contrasts and conditionals in consumer reviews. In Alexandra B, Erik VDG, Piek V, Andrés M, editors, 6th Workshop on Computational Approaches to Subjectivity, Sentiment & Social Media Analysis, WASSA 2015 : Workshop proceedings. Association for Computational Linguistics. 2015. p. 162-168

Detecting speculations, contrasts and conditionals in consumer reviews

Abstract

Conference

Subject classification (UKÄ)

Free keywords

Access to Document

Fingerprint

Projects

StaViCTA - Advances in the description and explanation of stance in discourse using visual and computational text analytics

Cite this