Polarity and intensity: the two aspects of sentiment analysis

Leimin Tian, Catherine Lai, Johanna D. Moore

Research output: Chapter in Book/Report/Conference proceedingConference PaperResearch

Abstract

Current multimodal sentiment analysis frames sentiment score prediction as a general Machine Learning task. However, what the sentiment score actually represents has often been overlooked. As a measurement of opinions and affective states, a sentiment score generally consists of two aspects: polarity and intensity. We decompose sentiment scores into these two aspects and study how they are conveyed through individual modalities and combined multimodal models in a naturalistic monologue setting. In particular, we build unimodal and multimodal multitask learning models with sentiment score prediction as the main task and polarity and/or intensity classification as the auxiliary tasks. Our experiments show that sentiment analysis benefits from multi-task learning, and individual modalities differ when conveying the polarity and intensity aspects of sentiment.
Original languageEnglish
Title of host publicationACL 2018 - First Grand Challenge and Workshop on Human Multimodal Language (Challenge-HML)
Subtitle of host publicationProceedings of the Workshop - July 20, 2018 Melbourne, Australia
EditorsAmir Zadeh, Louis-Philippe Morency, Paul Pu Liang, Soujanya Poria, Erik Cambria, Stefan Scherer
Place of PublicationStroudsburg PA USA
PublisherAssociation for Computational Linguistics (ACL)
Pages40-47
Number of pages8
ISBN (Electronic)9781948087469
Publication statusPublished - 2018
Externally publishedYes
EventGrand Challenge and Workshopon Human Multimodal Language 2018 - Melbourne, Australia
Duration: 20 Jul 201820 Jul 2018
http://multicomp.cs.cmu.edu/acl2018multimodalchallenge/

Conference

ConferenceGrand Challenge and Workshopon Human Multimodal Language 2018
Abbreviated titleChallenge-HML 2018
CountryAustralia
CityMelbourne
Period20/07/1820/07/18
Internet address

Cite this

Tian, L., Lai, C., & Moore, J. D. (2018). Polarity and intensity: the two aspects of sentiment analysis. In A. Zadeh, L-P. Morency, P. Pu Liang, S. Poria, E. Cambria, & S. Scherer (Eds.), ACL 2018 - First Grand Challenge and Workshop on Human Multimodal Language (Challenge-HML): Proceedings of the Workshop - July 20, 2018 Melbourne, Australia (pp. 40-47). Stroudsburg PA USA: Association for Computational Linguistics (ACL).
Tian, Leimin ; Lai, Catherine ; Moore, Johanna D. / Polarity and intensity : the two aspects of sentiment analysis. ACL 2018 - First Grand Challenge and Workshop on Human Multimodal Language (Challenge-HML): Proceedings of the Workshop - July 20, 2018 Melbourne, Australia. editor / Amir Zadeh ; Louis-Philippe Morency ; Paul Pu Liang ; Soujanya Poria ; Erik Cambria ; Stefan Scherer. Stroudsburg PA USA : Association for Computational Linguistics (ACL), 2018. pp. 40-47
@inproceedings{6eccfc5e2ad24e4ca789078328fd12e8,
title = "Polarity and intensity: the two aspects of sentiment analysis",
abstract = "Current multimodal sentiment analysis frames sentiment score prediction as a general Machine Learning task. However, what the sentiment score actually represents has often been overlooked. As a measurement of opinions and affective states, a sentiment score generally consists of two aspects: polarity and intensity. We decompose sentiment scores into these two aspects and study how they are conveyed through individual modalities and combined multimodal models in a naturalistic monologue setting. In particular, we build unimodal and multimodal multitask learning models with sentiment score prediction as the main task and polarity and/or intensity classification as the auxiliary tasks. Our experiments show that sentiment analysis benefits from multi-task learning, and individual modalities differ when conveying the polarity and intensity aspects of sentiment.",
author = "Leimin Tian and Catherine Lai and Moore, {Johanna D.}",
year = "2018",
language = "English",
pages = "40--47",
editor = "Zadeh, {Amir } and Morency, {Louis-Philippe } and {Pu Liang}, {Paul } and Poria, {Soujanya } and Cambria, {Erik } and Scherer, {Stefan }",
booktitle = "ACL 2018 - First Grand Challenge and Workshop on Human Multimodal Language (Challenge-HML)",
publisher = "Association for Computational Linguistics (ACL)",

}

Tian, L, Lai, C & Moore, JD 2018, Polarity and intensity: the two aspects of sentiment analysis. in A Zadeh, L-P Morency, P Pu Liang, S Poria, E Cambria & S Scherer (eds), ACL 2018 - First Grand Challenge and Workshop on Human Multimodal Language (Challenge-HML): Proceedings of the Workshop - July 20, 2018 Melbourne, Australia. Association for Computational Linguistics (ACL), Stroudsburg PA USA, pp. 40-47, Grand Challenge and Workshopon Human Multimodal Language 2018, Melbourne, Australia, 20/07/18.

Polarity and intensity : the two aspects of sentiment analysis. / Tian, Leimin; Lai, Catherine; Moore, Johanna D.

ACL 2018 - First Grand Challenge and Workshop on Human Multimodal Language (Challenge-HML): Proceedings of the Workshop - July 20, 2018 Melbourne, Australia. ed. / Amir Zadeh; Louis-Philippe Morency; Paul Pu Liang; Soujanya Poria; Erik Cambria; Stefan Scherer. Stroudsburg PA USA : Association for Computational Linguistics (ACL), 2018. p. 40-47.

Research output: Chapter in Book/Report/Conference proceedingConference PaperResearch

TY - GEN

T1 - Polarity and intensity

T2 - the two aspects of sentiment analysis

AU - Tian, Leimin

AU - Lai, Catherine

AU - Moore, Johanna D.

PY - 2018

Y1 - 2018

N2 - Current multimodal sentiment analysis frames sentiment score prediction as a general Machine Learning task. However, what the sentiment score actually represents has often been overlooked. As a measurement of opinions and affective states, a sentiment score generally consists of two aspects: polarity and intensity. We decompose sentiment scores into these two aspects and study how they are conveyed through individual modalities and combined multimodal models in a naturalistic monologue setting. In particular, we build unimodal and multimodal multitask learning models with sentiment score prediction as the main task and polarity and/or intensity classification as the auxiliary tasks. Our experiments show that sentiment analysis benefits from multi-task learning, and individual modalities differ when conveying the polarity and intensity aspects of sentiment.

AB - Current multimodal sentiment analysis frames sentiment score prediction as a general Machine Learning task. However, what the sentiment score actually represents has often been overlooked. As a measurement of opinions and affective states, a sentiment score generally consists of two aspects: polarity and intensity. We decompose sentiment scores into these two aspects and study how they are conveyed through individual modalities and combined multimodal models in a naturalistic monologue setting. In particular, we build unimodal and multimodal multitask learning models with sentiment score prediction as the main task and polarity and/or intensity classification as the auxiliary tasks. Our experiments show that sentiment analysis benefits from multi-task learning, and individual modalities differ when conveying the polarity and intensity aspects of sentiment.

M3 - Conference Paper

SP - 40

EP - 47

BT - ACL 2018 - First Grand Challenge and Workshop on Human Multimodal Language (Challenge-HML)

A2 - Zadeh, Amir

A2 - Morency, Louis-Philippe

A2 - Pu Liang, Paul

A2 - Poria, Soujanya

A2 - Cambria, Erik

A2 - Scherer, Stefan

PB - Association for Computational Linguistics (ACL)

CY - Stroudsburg PA USA

ER -

Tian L, Lai C, Moore JD. Polarity and intensity: the two aspects of sentiment analysis. In Zadeh A, Morency L-P, Pu Liang P, Poria S, Cambria E, Scherer S, editors, ACL 2018 - First Grand Challenge and Workshop on Human Multimodal Language (Challenge-HML): Proceedings of the Workshop - July 20, 2018 Melbourne, Australia. Stroudsburg PA USA: Association for Computational Linguistics (ACL). 2018. p. 40-47