Practical considerations on the use of preference learning for ranking emotional speech

Lotfian, Reza; Busso, Carlos

doi:10.1109/icassp.2016.7472670

Cited by 28 publications

(21 citation statements)

References 19 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Several studies have transformed values of affect to ordered ranks and then derived affect models via preference learning. As we will see in Section 4 such a transformation improves cross-validation capacities [60], [71], [72].…”

Section: Annotations Are Interval (Not Ordinal)mentioning

confidence: 84%

“…Recent work in speech-based affect recognition has demonstrated the benefits of using preference learning with ordinal labels [71], [72], [73], [93], [127]. Using time-continuous evaluations for arousal and valence provided by FeelTrace, the above studies defined preferences between pairs of speech samples and compared preference learning (via RankSVM) against binary classification and regression for modeling arousal and valence.…”

Section: Speechmentioning

confidence: 99%

“…The blank cells of the table call for further studies and analysis. [60], [71], [72], [73], [93] Music/Sounds Reliability [70], [110] Validity [70], [110] [118] Games…”

Section: A Summarymentioning

confidence: 99%

“…Case studies on speech: (a) improved precision at K (P@K) of RankSVM over regression (SVR) and binary classification (SVM) for arousal[71], (b) QA framework to identify trends from emotion annotation traces[72].…”

mentioning

confidence: 99%

See 3 more Smart Citations

The Ordinal Nature of Emotions: An Emerging Approach

Yannakakis

Cowie

Busso

2021

IEEE Trans. Affective Comput.

Self Cite

View full text Add to dashboard Cite

Computational representation of everyday emotional states is a challenging task and, arguably, one of the most fundamental for affective computing. Standard practice in emotion annotation is to ask people to assign a value of intensity or a class value to each emotional behavior they observe. Psychological theories and evidence from multiple disciplines including neuroscience, economics and artificial intelligence, however, suggest that the task of assigning reference-based values to subjective notions is better aligned with the underlying representations.This paper draws together the theoretical reasons to favor ordinal labels for representing and annotating emotion, reviewing the literature across several disciplines. We go on to discuss good and bad practices of treating ordinal and other forms of annotation data and make the case for preference learning methods as the appropriate approach for treating ordinal labels. We finally discuss the advantages of ordinal annotation with respect to both reliability and validity through a number of case studies in affective computing, and address common objections to the use of ordinal data. More broadly, the thesis that emotions are by nature ordinal is supported by both theoretical arguments and evidence, and opens new horizons for the way emotions are viewed, represented and analyzed computationally.

show abstract

Section: Annotations Are Interval (Not Ordinal)mentioning

confidence: 84%

Section: Speechmentioning

confidence: 99%

“…The blank cells of the table call for further studies and analysis. [60], [71], [72], [73], [93] Music/Sounds Reliability [70], [110] Validity [70], [110] [118] Games…”

Section: A Summarymentioning

confidence: 99%

mentioning

confidence: 99%

See 2 more Smart Citations

The Ordinal Nature of Emotions: An Emerging Approach

Yannakakis

Cowie

Busso

2021

IEEE Trans. Affective Comput.

Self Cite

View full text Add to dashboard Cite

show abstract

“…Rank Support Vector Machines (RankSVM), a variant of SVMs, was introduced by Joachims [12] as a way of ranking webpages based on their click rate. A RankSVM consists of projecting pairwise data onto a feature space combined with ranked annotations, adjusting a weight vector ( w) so that all points in the training dataset are ordered by their projection onto w. Although RankSVMs started as a way of optimizing webpage queries, it has been applied to several other domains quite successfully such as for the detection of emotion in speech [36] and musical pieces [37].…”

Section: Preference Learning For Affect Modellingmentioning

confidence: 99%

Modelling Affect for Horror Soundscapes

Lopes

Liapis

Yannakakis

2019

IEEE Trans. Affective Comput.

View full text Add to dashboard Cite

The feeling of horror within movies or games relies on the audience's perception of a tense atmosphere-often achieved through sound accompanied by the on-screen drama-guiding its emotional experience throughout the scene or game-play sequence. These progressions are often crafted through an a priori knowledge of how a scene or game-play sequence will playout, and the intended emotional patterns a game director wants to transmit. The appropriate design of sound becomes even more challenging once the scenery and the general context is autonomously generated by an algorithm. Towards realizing sound-based affective interaction in games this paper explores the creation of computational models capable of ranking short audio pieces based on crowdsourced annotations of tension, arousal and valence. Affect models are trained via preference learning on over a thousand annotations with the use of support vector machines, whose inputs are low-level features extracted from the audio assets of a comprehensive sound library. The models constructed in this work are able to predict the tension, arousal and valence elicited by sound, respectively, with an accuracy of approximately 65%, 66% and 72%.

show abstract