Gabriel P. Oliveira

The viral phenomenon is present in several contexts, combining the advantages of streaming platforms and other social networks. Music is no exception. Viral songs are widely shared in a short amount of time, and they may become successful by reaching the top of the charts with millions of streams and digital sales. However, not all songs that go viral become hits, as the sharing process is not enough to be converted into streams. In this work, we analyze viral and hit songs as two different yet interconnected aspects of music popularity. Specifically, we aim to uncover factors that are relevant to distinguishing viral from hit songs. We evaluate three research hypotheses related to musical features, and the results reveal that considering only acoustic and other intrinsic features is not enough; and extrinsic features (e.g., artists’ genre and the time from the release to the first chart entry) are essential in achieving such a goal. Moreover, artist-related and temporal features are among the most relevant indicators to differentiate hit and viral songs. Overall, our findings offer relevant insights for understanding the dynamics of music consumption and its sharing on online platforms, as we reveal important factors that describe what makes a viral song and what differentiates it from a hit.

Hit Song Science (HSS) is an emerging topic that aims to unveil the success dynamics within the music industry. Considering the growth of the area, we provide a comprehensive study with a complete review of the main topics of this interdisciplinary field from a computer science perspective. We also define a generic workflow for HSS, introduce taxonomies for success measures and musical features, and categorize the main current learning algorithms. Overall, this survey may serve as a starting point for future research on HSS, as it emerges as a promising field that benefits both the academy and the music industry.

In this work, we reveal fundamental patterns that appear in individual musical careers. Such careers may go through ups and downs depending on the current market moment and release of new songs. In particular, they face hot streak periods in which high-impact bursts occur in sequence. Identifying such periods and even predicting them may help in other practical issues, which include foreseeing success and recommending artists. After modeling artists’ careers as time series, we find a general trend of clustering within the most successful weeks, which justifies the applicability of the concept of hot streaks. Hence, we use a specific methodology for identifying hot streaks, whose evaluation results reveal meaningful patterns for artists of different genres. We also confirm the career peaks of artists appear and disappear progressively over time. Overall, our findings shed light on the science of musical success as we observe the temporal evolution of artists’ careers and their hot streaks.

Assessing the collaboration among developers is important to understand different aspects of software lifecycle including code smell intensity, bug fixes, and software quality. This kind of collaboration can be obtained from social networks, which represent interactions between individuals in different contexts. In this paper, we model GitHub developers’ collaborations in a heterogeneous network by considering three aspects: social collaboration, collaboration time in a repository and technical features. Then, we explore the GitHub network from different perspectives: size, relevance, and potential applications. The results show the considered metrics are not correlated, bringing new information about the collaborations. We also show that such information is useful for social developer ranking, an actual task which is often part of different applications, such as team formation, community detection and pair programming. Finally, as software quality is intrinsic to the people who code it, our methodology and analyses represent initial steps towards people-centered software quality analysis, as further discussed throughout this article.

We analyze and identify collaboration profiles in success-based music genre networks. Such networks are built upon data recently collected from both global and regional Spotify weekly charts. Overall, our findings reveal an increase in the number of distinct successful genres from high-potential markets, pointing out that local repertoire is more important than ever on building the global music ecosystem. We also detect collaboration patterns mapped into four different profiles: Solid, Regular, Bridge and Emerging, wherein the two first depict higher average success. These findings indicate great opportunities for the music industry by revealing the driving power of inter-genre collaborations within regional and global markets.