Standard

Detecting Socially Significant Music Events using Temporally Noisy Labels. / Yadati, Karthik; Larson, Martha; Liem, Cynthia C.S.; Hanjalic, Alan.

In: IEEE Transactions on Multimedia, Vol. 20, No. 9, 2018, p. 2526-2540.

Research output: Contribution to journalArticleScientificpeer-review

Harvard

APA

Vancouver

Author

BibTeX

@article{df31b873e0d84a33aac1e098f133f04e,
title = "Detecting Socially Significant Music Events using Temporally Noisy Labels",
abstract = "In this paper, we focus on event detection over the timeline of a music track. Such technology is motivated by the need for innovative applications such as searching, non-linearaccess and recommendation. Event detection over the timeline requires time-code level labels in order to train machine learning dels. We use timed comments from SoundCloud, a modern social music sharing platform, to obtain these labels. While in this way the need for tedious and time-consuming manual labeling can be reduced, the challenge is that timed comments are subject to additive temporal noise, as they are in the temporal neighborhood of the actual events. We investigate the utility of such noisy timed comments as training labels through a case study, in which we investigate three types of events in Electronic Dance Music (EDM): drop, build and break. These socially significant events play a key role in an EDM track's unfolding and are popular in social media circles. They are therefore not only interesting for detection, but also typically accompanied by timed comments resulting from the online social activity around them. We propose a two-stage learning method that relies on noisy timed comments and, given a music track, marks the events on the timeline. In the experiments, we focus in particular on investigating to which extent noisy timed comments can replace manually added expert labels. The conclusions we draw during this study provide useful insights that motivates further research in the field of event detection.",
keywords = "break, build, drop, EDM, event, Event detection, Feature extraction, Music, Noise measurement, SoundCloud, Streaming media, timed comments, Training",
author = "Karthik Yadati and Martha Larson and Liem, {Cynthia C.S.} and Alan Hanjalic",
note = "Green Open Access added to TU Delft Institutional Repository ‘You share, we take care!’ – Taverne project https://www.openaccess.nl/en/you-share-we-take-care Otherwise as indicated in the copyright section: the publisher is the copyright holder of this work and the author uses the Dutch legislation to make this work public.",
year = "2018",
doi = "10.1109/TMM.2018.2801719",
language = "English",
volume = "20",
pages = "2526--2540",
journal = "IEEE Transactions on Multimedia",
issn = "1520-9210",
publisher = "IEEE",
number = "9",

}

RIS

TY - JOUR

T1 - Detecting Socially Significant Music Events using Temporally Noisy Labels

AU - Yadati, Karthik

AU - Larson, Martha

AU - Liem, Cynthia C.S.

AU - Hanjalic, Alan

N1 - Green Open Access added to TU Delft Institutional Repository ‘You share, we take care!’ – Taverne project https://www.openaccess.nl/en/you-share-we-take-care Otherwise as indicated in the copyright section: the publisher is the copyright holder of this work and the author uses the Dutch legislation to make this work public.

PY - 2018

Y1 - 2018

N2 - In this paper, we focus on event detection over the timeline of a music track. Such technology is motivated by the need for innovative applications such as searching, non-linearaccess and recommendation. Event detection over the timeline requires time-code level labels in order to train machine learning dels. We use timed comments from SoundCloud, a modern social music sharing platform, to obtain these labels. While in this way the need for tedious and time-consuming manual labeling can be reduced, the challenge is that timed comments are subject to additive temporal noise, as they are in the temporal neighborhood of the actual events. We investigate the utility of such noisy timed comments as training labels through a case study, in which we investigate three types of events in Electronic Dance Music (EDM): drop, build and break. These socially significant events play a key role in an EDM track's unfolding and are popular in social media circles. They are therefore not only interesting for detection, but also typically accompanied by timed comments resulting from the online social activity around them. We propose a two-stage learning method that relies on noisy timed comments and, given a music track, marks the events on the timeline. In the experiments, we focus in particular on investigating to which extent noisy timed comments can replace manually added expert labels. The conclusions we draw during this study provide useful insights that motivates further research in the field of event detection.

AB - In this paper, we focus on event detection over the timeline of a music track. Such technology is motivated by the need for innovative applications such as searching, non-linearaccess and recommendation. Event detection over the timeline requires time-code level labels in order to train machine learning dels. We use timed comments from SoundCloud, a modern social music sharing platform, to obtain these labels. While in this way the need for tedious and time-consuming manual labeling can be reduced, the challenge is that timed comments are subject to additive temporal noise, as they are in the temporal neighborhood of the actual events. We investigate the utility of such noisy timed comments as training labels through a case study, in which we investigate three types of events in Electronic Dance Music (EDM): drop, build and break. These socially significant events play a key role in an EDM track's unfolding and are popular in social media circles. They are therefore not only interesting for detection, but also typically accompanied by timed comments resulting from the online social activity around them. We propose a two-stage learning method that relies on noisy timed comments and, given a music track, marks the events on the timeline. In the experiments, we focus in particular on investigating to which extent noisy timed comments can replace manually added expert labels. The conclusions we draw during this study provide useful insights that motivates further research in the field of event detection.

KW - break

KW - build

KW - drop

KW - EDM

KW - event

KW - Event detection

KW - Feature extraction

KW - Music

KW - Noise measurement

KW - SoundCloud

KW - Streaming media

KW - timed comments

KW - Training

UR - http://www.scopus.com/inward/record.url?scp=85041648412&partnerID=8YFLogxK

U2 - 10.1109/TMM.2018.2801719

DO - 10.1109/TMM.2018.2801719

M3 - Article

VL - 20

SP - 2526

EP - 2540

JO - IEEE Transactions on Multimedia

T2 - IEEE Transactions on Multimedia

JF - IEEE Transactions on Multimedia

SN - 1520-9210

IS - 9

ER -

ID: 44184733