Sources |
After compiling our sources and looking at research on the musical industry in the 20th and 21st century to complement our main dataset about the Billboard Top 100, we wanted to look at the question about how the emergency of the Internet and different technologies impacted the diversity of popular music impacted the diversity of popular songs on the Billboard Hot 100. We defined diversity initially as the number of genres tagged in a year for popular music. We pinpointed the Internet as 1995, as discussed in class, as well as Philip Hayward’s article “Enterprise on the new Frontier”. We thought about the construction of the Billboard Hot 100, and how it is interconnected with radio and other mediums of distributing music. We also thought about the relationship between power, data, and narrative, from the Michel-Rolph Trouillot reading we explored in class, making us wonder about who isn’t represented in the Billboard and why. We used Google Scholar and JSTOR, library databases to find relevant texts based on metadata.
|
Processing |
Processing, the second part of the digital humanities project was a very intensive part of our project. We cleaned data using R and Excel because Tableau could not handle such a large dataset. Moreover, we used data visualization in Excel and R to see trends that were easier to pick up through visual exploration than by looking at long spreadsheets. Through visualizing the data, we developed insights on relationships between time and genres. This is a reminder of Yau’s definition of data visualization as “a medium: a way to explore, present, and express meaning in data” (Yau, 44). We also used the secondary dataset (linked to the main Billboard dataset) to drive most of the research for better interpretation of diversity and creating our statistical methods to aggregate and define the changes in the music industry due to advent of Internet. The secondary dataset was a crowd-sourced dataset that allowed many users to tag songs with a range of tags that are “non-standard”, including “progressive rock” and “post-punk”. We also filtered out genres with less than 50 tags for most of the data visualizations as we felt that there were not enough data points to represent this kind of popular music, being more obscure in description. Finally, our group sought ways to quantitatively interpret very qualitative aspects of the data, and our pre-defined statistical methods were probably most helpful in this endeavor.
|
Presentation |
Finally, the Popular Music project culminated in a presentation stage, where we distilled major trends and represented what we learnt from our data in an interactive website. It reflected the strides we made to overcome steep learning curves as humanists grappling with computational tools. After much initial exploratory data analysis, we used Tableau to make many of our data visualizations because it is interactive, allowing people to play around with the data visually and follow trends in more complex graphs. A genre we found particularly significant was Rock, a musical genre dominated by white males, which decreased a lot in popularity over the years. Hip-hop increased a lot since 1995, which tends to have more artists of color, hence showing an increase in diversity. We also found that the number of distinct artists increased over time, also indicating an increase in the diversity of the music that becomes popular in sales, even though platforms like Napster and Limewire disrupted the music industry through peer-to-peer networks. This is also related to the number of people accessing the Internet during this period; there is a huge increase as it became more affordable and accessible. The discussion of lexical frequency profiling in the Archer reading from our DH101 class informed the creative representation of genres on geographical maps. We overlaid genre counts before and after the Internet (1995) on the US map using Voyant Tools, which helps in relaying the idea of an American social construct - the Billboard Top 100. This demonstrates what was more popular before and after in terms of genre. The timeline provides historical background when analyzing our data. This allows us to consider the variations of chart types and the various metrics used when calculating song positioning. With this information, we are better able to analyze the accuracy of the data and situate it within the context of our project.
|