Corpus Analysis Tools For Computational Hook Discovery.

Corpus Analysis Tools For Computational Hook Discovery. Balen, J. V., Burgoyne, J. A., Bountouridis, D., Müllensiefen, D., & Veltkamp, R. C. October, 2015. Publisher: Zenodo

Paper doi abstract bibtex

Compared to studies with symbolic music data, advances in music description from audio have overwhelmingly focused on ground truth reconstruction and maximizing prediction accuracy, with only a small fraction of studies using audio description to gain insight into musical data. We present a strategy for the corpus analysis of audio data that is optimized for interpretable results. The approach brings two previously unexplored concepts to the audio domain: audio bigram distributions, and the use of corpus-relative or “second-order” descriptors. To test the real-world applicability of our method, we present an experiment in which we model song recognition data collected in a widely-played music game. By using the proposed corpus analysis pipeline we are able to present a cognitively adequate analysis that allows a model interpretation in terms of the listening history and experience of our participants. We ﬁnd that our corpus-based audio features are able to explain a comparable amount of variance to symbolic features for this task when used alone and that they can supplement symbolic features proﬁtably when the two types of features are used in tandem. Finally, we highlight new insights into what makes music recognizable.

@Article{          balen.ea2015-corpus,
    author       = {Balen, Jan Van and Burgoyne, John Ashley and
                   Bountouridis, Dimitrios and Müllensiefen, Daniel and
                   Veltkamp, Remco C.},
    year         = {2015},
    title        = {Corpus {Analysis} {Tools} {For} {Computational} {Hook}
                   {Discovery}.},
    copyright    = {Creative Commons Attribution 4.0, Open Access},
    url          = {https://zenodo.org/record/1415038},
    doi          = {10.5281/ZENODO.1415038},
    abstract     = {Compared to studies with symbolic music data, advances in
                   music description from audio have overwhelmingly focused
                   on ground truth reconstruction and maximizing prediction
                   accuracy, with only a small fraction of studies using
                   audio description to gain insight into musical data. We
                   present a strategy for the corpus analysis of audio data
                   that is optimized for interpretable results. The approach
                   brings two previously unexplored concepts to the audio
                   domain: audio bigram distributions, and the use of
                   corpus-relative or “second-order” descriptors. To test
                   the real-world applicability of our method, we present an
                   experiment in which we model song recognition data
                   collected in a widely-played music game. By using the
                   proposed corpus analysis pipeline we are able to present a
                   cognitively adequate analysis that allows a model
                   interpretation in terms of the listening history and
                   experience of our participants. We ﬁnd that our
                   corpus-based audio features are able to explain a
                   comparable amount of variance to symbolic features for
                   this task when used alone and that they can supplement
                   symbolic features proﬁtably when the two types of
                   features are used in tandem. Finally, we highlight new
                   insights into what makes music recognizable.},
    language     = {en},
    urldate      = {2023-02-23},
    month        = oct,
    note         = {Publisher: Zenodo}
}

Downloads: 0

{"_id":"jGyE9DkRX3ptNADM6","bibbaseid":"balen-burgoyne-bountouridis-mllensiefen-veltkamp-corpusanalysistoolsforcomputationalhookdiscovery-2015","author_short":["Balen, J. V.","Burgoyne, J. A.","Bountouridis, D.","Müllensiefen, D.","Veltkamp, R. C."],"bibdata":{"bibtype":"article","type":"article","author":[{"propositions":[],"lastnames":["Balen"],"firstnames":["Jan","Van"],"suffixes":[]},{"propositions":[],"lastnames":["Burgoyne"],"firstnames":["John","Ashley"],"suffixes":[]},{"propositions":[],"lastnames":["Bountouridis"],"firstnames":["Dimitrios"],"suffixes":[]},{"propositions":[],"lastnames":["Müllensiefen"],"firstnames":["Daniel"],"suffixes":[]},{"propositions":[],"lastnames":["Veltkamp"],"firstnames":["Remco","C."],"suffixes":[]}],"year":"2015","title":"Corpus Analysis Tools For Computational Hook Discovery.","copyright":"Creative Commons Attribution 4.0, Open Access","url":"https://zenodo.org/record/1415038","doi":"10.5281/ZENODO.1415038","abstract":"Compared to studies with symbolic music data, advances in music description from audio have overwhelmingly focused on ground truth reconstruction and maximizing prediction accuracy, with only a small fraction of studies using audio description to gain insight into musical data. We present a strategy for the corpus analysis of audio data that is optimized for interpretable results. The approach brings two previously unexplored concepts to the audio domain: audio bigram distributions, and the use of corpus-relative or “second-order” descriptors. To test the real-world applicability of our method, we present an experiment in which we model song recognition data collected in a widely-played music game. By using the proposed corpus analysis pipeline we are able to present a cognitively adequate analysis that allows a model interpretation in terms of the listening history and experience of our participants. We ﬁnd that our corpus-based audio features are able to explain a comparable amount of variance to symbolic features for this task when used alone and that they can supplement symbolic features proﬁtably when the two types of features are used in tandem. Finally, we highlight new insights into what makes music recognizable.","language":"en","urldate":"2023-02-23","month":"October","note":"Publisher: Zenodo","bibtex":"@Article{ balen.ea2015-corpus,\n author = {Balen, Jan Van and Burgoyne, John Ashley and\n Bountouridis, Dimitrios and Müllensiefen, Daniel and\n Veltkamp, Remco C.},\n year = {2015},\n title = {Corpus {Analysis} {Tools} {For} {Computational} {Hook}\n {Discovery}.},\n copyright = {Creative Commons Attribution 4.0, Open Access},\n url = {https://zenodo.org/record/1415038},\n doi = {10.5281/ZENODO.1415038},\n abstract = {Compared to studies with symbolic music data, advances in\n music description from audio have overwhelmingly focused\n on ground truth reconstruction and maximizing prediction\n accuracy, with only a small fraction of studies using\n audio description to gain insight into musical data. We\n present a strategy for the corpus analysis of audio data\n that is optimized for interpretable results. The approach\n brings two previously unexplored concepts to the audio\n domain: audio bigram distributions, and the use of\n corpus-relative or “second-order” descriptors. To test\n the real-world applicability of our method, we present an\n experiment in which we model song recognition data\n collected in a widely-played music game. By using the\n proposed corpus analysis pipeline we are able to present a\n cognitively adequate analysis that allows a model\n interpretation in terms of the listening history and\n experience of our participants. We ﬁnd that our\n corpus-based audio features are able to explain a\n comparable amount of variance to symbolic features for\n this task when used alone and that they can supplement\n symbolic features proﬁtably when the two types of\n features are used in tandem. Finally, we highlight new\n insights into what makes music recognizable.},\n language = {en},\n urldate = {2023-02-23},\n month = oct,\n note = {Publisher: Zenodo}\n}\n\n","author_short":["Balen, J. V.","Burgoyne, J. A.","Bountouridis, D.","Müllensiefen, D.","Veltkamp, R. C."],"key":"balen.ea2015-corpus","id":"balen.ea2015-corpus","bibbaseid":"balen-burgoyne-bountouridis-mllensiefen-veltkamp-corpusanalysistoolsforcomputationalhookdiscovery-2015","role":"author","urls":{"Paper":"https://zenodo.org/record/1415038"},"metadata":{"authorlinks":{}}},"bibtype":"article","biburl":"https://hmb.sampaio.me/bibliografia.bib.txt","dataSources":["n6MFY2CscQLDpJ7nT"],"keywords":[],"search_terms":["corpus","analysis","tools","computational","hook","discovery","balen","burgoyne","bountouridis","müllensiefen","veltkamp"],"title":"Corpus Analysis Tools For Computational Hook Discovery.","year":2015}