Learning to Recognize Musical Genre from Audio

Defferrard, Michaël; Mohanty, Sharada P.; Carroll, Sean F.; Salathé, Marcel

doi:10.1145/3184558.3192310

Cited by 13 publications

(6 citation statements)

References 3 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…5c, we show that our proposed MAM is able to recover the missing framework by pre-training over Librispeech dataset. Since MAM does not need any transcription to perform pre-training, we also pre-train MAM with FMA corpus (Defferrard et al, 2018), which is a music dataset. Surprisingly, MAM performs very similar reconstruction ability compared with the one that are pre-trained with speech dataset considering the corrupted audio is only about speech.…”

Section: Methodsmentioning

confidence: 99%

MAM: Masked Acoustic Modeling for End-to-End Speech-to-Text Translation

Chen,

Ma,

Zheng

et al. 2020

Preprint

View full text Add to dashboard Cite

End-to-end Speech-to-text Translation (E2E-ST), which directly translates source language speech to target language text, is widely useful in practice, but traditional cascaded approaches (ASR+MT) often suffer from error propagation in the pipeline. On the other hand, existing end-to-end solutions heavily depend on the source language transcriptions for pre-training or multi-task training with Automatic Speech Recognition (ASR). We instead propose a simple technique to learn a robust speech encoder in a self-supervised fashion only on the speech side, which can utilize speech data without transcription. This technique, termed Masked Acoustic Modeling (MAM), can also perform pre-training, for the first time, on any acoustic signals (including non-speech ones) without annotation. Compared with current state-of-the-art models on ST, our technique achieves +1.4 BLEU improvement without using transcriptions, and +1.2 BLEU using transcriptions. Pre-training of MAM with arbitrary acoustic signals also boosts the downstream speech-related tasks.

show abstract

Section: Methodsmentioning

confidence: 99%

MAM: Masked Acoustic Modeling for End-to-End Speech-to-Text Translation

Chen,

Ma,

Zheng

et al. 2020

Preprint

View full text Add to dashboard Cite

show abstract

“…We chose to use the FMA dataset [37], an open dataset that allows reproducibility and comparison of future work. Due to size constraints, we only consider the medium split that includes 25.000 music tracks spread into 16 genres.…”

Section: Considered Dataset and Deepfake Generatorsmentioning

confidence: 99%

Of Spiky SVDs and Music Recommendation

Afchar,

Hennequin,

Guigue

2023

Proceedings of the 17th ACM Conference on Recommender Systems

View full text Add to dashboard Cite

In the face of a new era of generative models, the detection of artificially generated content has become a matter of utmost importance. The ability to create credible minute-long music deepfakes in a few seconds on userfriendly platforms poses a real threat of fraud on streaming services and unfair competition to human artists. This paper demonstrates the possibility (and surprising ease) of training classifiers on datasets comprising real audio and fake reconstructions, achieving a convincing accuracy of 99.8%. To our knowledge, this marks the first publication of a music deepfake detector, a tool that will help in the regulation of music forgery. Nevertheless, informed by decades of literature on forgery detection in other fields, we stress that a good test score is not the end of the story. We step back from the straightforward ML framework and expose many facets that could be problematic with such a deployed detector: calibration, robustness to audio manipulation, generalisation to unseen models, interpretability and possibility for recourse. This second part acts as a position for future research steps in the field and a caveat to a flourishing market of fake content checkers.

show abstract

“…The most notable of such datasets is the FMA (Free Music Archive) [41]. A description of such dataset and preliminary analyses are described in [42].…”

Section: Lyrics Analysismentioning

confidence: 99%

Analysis of the Arab Singer Shadia’s Lyrics

Khamis

Gomaa

2023

Preprint

View full text Add to dashboard Cite

In this work, we analyze the lyrics of one of the most famous and influential Arab artists in the twentieth century, namely, \<شادية> (Shadia). Lyrics analysis provides a deep insight into the artist's career evolution, her interactions with the surrounding environment including the social, political, and economic conditions. In order to perform such analysis we had to collect and compile the lyrics of Shadia accompanied with the necessary metadata into an organized and structured form. The data are preprocessed by removing stop words and doing some normalization operations over the songs prose. We did not perform any lemmatization or stemming as the original form of the tokens convey much more information than the source words.We performed a lexical analysis in order to study both the lexical density and diversity over the course of Shadia career life. We have as well studied the most significant words, idioms, and terms played in the songs using tools such as word clouds and more quantitative measures such as term frequency-inverse document frequency. We have divided the career life of Shadia into sub-decades of length \(5\) years and all analyses are done both in a yearly fashion and more coarsely over such sub-decades. We have found a strong correlation between our statistical analysis and the socio-political status in Egypt and the Arab world during that time.This is especially relevant knowing that Shadia is very much truly considered the daughter of the generation of the 1952 revolution in Egypt. The significance of Shadia and her lyrics stem essentially from being contemporaneous to radical changes in Egypt across all sectors including political (support of liberation movements across the world, and the conflict with Israel), and socio-economic (especially changing the social class structure in Egypt). We also investigated the potential effectiveness of PoS (Part of Speech) tagging in genre analysis and classification.

show abstract

Learning to Recognize Musical Genre from Audio

Abstract: We here summarize our experience running a challenge with open data for musical genre recognition. Those notes motivate the task and the challenge design, show some statistics about the submissions, and present the results.

Cited by 13 publications

References 3 publications

MAM: Masked Acoustic Modeling for End-to-End Speech-to-Text Translation

MAM: Masked Acoustic Modeling for End-to-End Speech-to-Text Translation

Of Spiky SVDs and Music Recommendation

Analysis of the Arab Singer Shadia’s Lyrics

Contact Info

Product

Resources

About