A Multimodal End-to-End Deep Learning Architecture for Music Popularity Prediction

Abstract

The continuous evolution of multimedia applications is fostering applied research in order to dynamically enhance the services provided by platforms such as Spotify, Lastfm, or Billboard. Thus, innovative methods for retrieving specific information from large volumes of data related with music arises as a potential challenge within the Music Information Retrieval (MIR) framework. Moreover, despite the existence of several musical-based datasets, there is still a lack of information to properly assess an accurate estimation of the impact or the popularity of a song within a platform. Furthermore, the aforementioned platforms measure the popularity in various manners, thus increasing the difficulties in performing generalized and comparable models. In this paper, the creation of SpotGenTrack Popularity Dataset (SPD) is presented as an alternative solution to existing datasets that will facilitate researchers when comparing and promoting their models. In addition, an innovative multimodal end-to-end Deep Learning architecture named as HitMusicNet is presented for predicting popularity in music recordings. Experiments conducted show that the proposed architecture outperforms previous studies in the State-of-the-Art by incorporating three main modalities to the analysis, such as audio, lyrics and meta-data as well as a preliminary compression stage via autoencoder to better the capability of the model when predicting the popularity.

Predicting Music Popularity Using Machine Learning Algorithm and Music Metrics Available in Spotify

“…In paper [6], authors try to predict the song popularity using acoustic features including MFCC features, but their predictions had room for improvement. An end-to-end deep learning architecture named Hit Music Net is presented in the paper [7], which gives better results than other machine learning algorithms. Though it cannot be very helpful for particular music streaming app like Spotify.…”

PPORTAL: Public Domain Portuguese-language Literature Dataset

“…AI-powered technology has accelerated dramatically over the past few years in creative industries, such as music, movies, and books. With so much digital information available, machine learning solutions have been developed to predict the next hit song [Martín-Gutiérrez et al 2020], the next blockbuster [Ahmad et al 2017], or even the next bestseller [Wang et al 2019]. Moreover, book recommendation systems have also been a hot application in ML.…”

LSTM-RPA: A Simple but Effective Long Sequence Prediction Algorithm for Music Popularity Prediction

“…In [5], the researchers used the frequency spectrogram images of music raw audio signals as the CNN model input feature to predict popularity music, and the accuracy of this model is 61% in the test set. Research [12] established HitMusicNet based on multimodal end-to-end Deep Learning (DL) architecture. HitMusicNet used autoencoder to compress high-dimensional features, which consist of audio, text lyrics, and meta-data to improve the accuracy of the model prediction.…”