Kaitlynn Li
Each year, millions of people worldwide watch the broadcasting of the Academy Awards ceremony. There, the Oscar for “Best Original Song” is awarded to the “best” movie song of the year, based on the votes of almost 6000 Academy members. Past studies have been conducted on the instrumentation, complexity, and range of music where analysis of those aspects are linearly compared with popularity of songs, but key objective elements that are strong enough to predict with have not been found. The results of this study confirm that Oscar winning songs have certain key elements of music that voters find more important or likable compared to others. The key elements include presence of half note drum beats, a male vocalist(s), piano in the beginning followed by strings instrumentation, relative longer duration, lack of lyric order of verse followed by chorus, lack of a combination of electric and non-electric instruments, and larger relative first week box office revenue. Based on these elements, the logistic regression model identifies 15 out of 20 Oscar winning songs from 2000 to 2019 and correctly predicts the 2016, 2017, and 2019 winning songs. This approach used to decode Oscar winning songs can be applied to music awards worldwide, identify unique music preferences for different ages, nationalities, etc., and identify the changes in music preferences throughout history. The findings can influence the modern composing practice (from writing the next top hit to using preferable drum beats in the military band) and help create impactful music that meets societal needs.
Comments
Very interesting topic and research. I really like your explanation and the figures used. Good job Kaitlynn! – Terrian Spurs
I need this to help me win next year’s Oscar pool! Nicely done, Kaitlynn. – Lane Mann
This is such an interesting topic! How did you determine what data to collect for each song? Were there any elements you initially tried but found to be statistically insignificant? – Emily
Hello Emily, Thank you so much for your comment! I utilized other research on music analysis to compile a list of elements I wanted to look for in each song. The categorical element instrumentation I actually initially tried and the majority was insignificant by themselves; however, when I combined them (e.g. electrical vs. classical) they showed to be significant to predicting the songs. For example, if you hear piano at the beginning of the song followed by strings, it is actually more likely to win the oscars! Best, Kaitlynn – Kaitlynn Li
This is such a unique topic. How did you arrive at it? Are you hoping to do future work that might use your model to create music, or are you interested more in the modeling and analysis side? – Rob Reichle
Hello Rob, Throughout high school I was heavily involved in marching band and full orchestra so music has always been ingrained in the projects I pursue! I arrived at this topic after reading multiple articles on music analysis, music theory, and music therapy that have been published in the past few years. Personally, I am not a great composer so I would be more inclined to do modeling and analysis within the industries; however, I would love to see this project come to life through a music composition knowing the statistically significant elements! Best, Kaitlynn – Kaitlynn Li