Music and Math
“Music and Math”
The amazing advances that we have seen in the power of computing have led to information technology being applied to many aspects of our daily lives. Often information technology is working behind the scenes. A nice example of this is (which I came across during my minor “Exploration of New Markets through Innovation” at EUR):
HIT SONG SCIENCE (HSS):
HSS is a part of “Music Information Retrieval”; which is the science of retrieving information from music.
The term Hit Song Science was coined and trademarked by Mike McCready the co-founder of HMI-polyphonic and later X-ray (these operate within the so called “hit counselling business”).
HSS entails using statistical, signal processing and machine learning methods to attempt predicting the commercial success (or as some call it: “mainstream potential”) of a song by looking at characteristics of other songs that have been a success in the past.
Early studies have claimed that this technology seems to be successful in predicting the commercial success of songs. However other studies refute these claims (see: “hit song science: not yet a science”). Nonetheless these firms still exist and are evaluating songs as we speak.
The implications of the adoption of this technology by the music industry are extensive. As this technology has become widely available (Polyphonic HMI and musicxray), more stakeholders (not only the biggest record labels) in the music industry are able to benefit from this technology: producers, small record companies and even individual artists can use HSS.
I have no problem whatsoever if record companies want to use this technology as a data mining tool to be able to predict the success of songs. I also have no problem if artists and record companies use it to find one another (artists finding record labels that fit their sound and vice versa). However, the creators of this technology are now boasting that it cannot only predict the potential of a song but it is also able to suggest improvements in songs. This to me is a little worrying: I am not that keen on this software tampering with music to optimize the chance that it will be a success. I feel that music tampered in this way may lose its human touch/feel/sound.
The controversial part of HSS is that we perceive music as something very “human” and I myself don’t think machines can replace human ears when it comes to music. However this sounds a lot like the argument people use to argue on machines being able to play chess.
Even if this technology becomes successful and widely adopted by the music industry, I believe the positives (data mining and artist/label coordination) outweigh the negatives (losing the “human sound” in certain songs). I am not worried that the quality of music in general might decline due to HSS; it might have an impact on pop music. However, I believe that excellent music will always still be created. There will always be artists that refuse to use such technologies, even if such artists were to remain relatively “underground”.
Since the inception of the internet: underground is not all that underground anymore (see “from niches to riches: anatomy of the long tail”).
On a last note:
I am very curious as to musical compositions made solely by machines, which is a possible future application of Music Information Retrieval (however this hasn’t produced anything noteworthy as of yet).
What do you think of HSS? Will it survive in the (pop-)music industry? Do you think it will improve music or maybe make it more generic? Do you think it will shift the focus from artists to individual songs if this technology were to be even more widely adopted?
https://en.wikipedia.org/wiki/Music_information_retrieval
https://en.wikipedia.org/wiki/Mike_McCready_(music_entrepreneur)
http://www.csl.sony.fr/downloads/papers/2008/pachet-08c.pdf
https://en.wikipedia.org/wiki/Platinum_Blue_Music_Intelligence
Author: Euclid Alexis Haralambidis
S.I.D.: 313081eh
4 responses to “Music and Math”
Leave a comment
Recent Comments
| From Farmville to th… on The success of Estonia’s… | |
 | Tycho van der Scheer on How will traditional grocers c… |
 | 374482rm on Bitcoin The Disruptor |
 | VH on Makeup Genius: How L’Oréal is… |
 | Josephine on 3D printing: a new standard of… |
Information Strategy 
This is an interesting post. A while back I read about Shazam’s capability to predict hit songs (http://www.factmag.com/2015/05/08/shazam-says-they-can-predict-a-1-hit-over-a-month-in-advance/).
But I have to agree that this data will have a big impact on pop music. You already see producers, artist, songwriters, label etc. looking at what sells and copying those songs. And when the song is made, they also give it alot of marketing by promoting it on radiostations etc. With this data, they will pinpoint new songs they can easily generate more revenue/exposure for. I believe this they could also apply for upcoming artist. Pop music is already generic, otherwise it wouldn’t be pop(ular).
What must be stated that original music will always be made. The question is, will it become mainstream and is that predictable?
Personally I also want to know from which country the data is being collected. Is it perhaps possible to pinpoint popular songs in city’s or countries? Or is it only worldwide?
First of all it was great to read about this technology, I never heard of it before and I am always amazed by these innovations so thanks a lot for writing about it 🙂 Secondly I really appreciate the perspective that you share in this article (“the positives outweigh the negatives” “good music will always be created”) because I have seen many times people “ringing alarms” about such technologies mentioning all the possible future negatives sides and painting a picture of (in these case music) dying out by this. And it is cool to see a perspective which mentions possible threats but focuses on the positive sides. So thanks it was a good and super interesting weekend morning read!
Hello,
My name is Ekaterina Marinova, and I really liked the article. This is the first time I hear about this, so I found it really informative and innovative.
However, I have few concerns about the technology. Are there any success criteria specified? I believe that one cannot categorize a song as successfull, since it varies among genres. For example, I believe that the probability of a song in the pop music is higher, compared to that of a song in jazz, if the measure criteria is the people who are going to listen to it, and download it. I also believe that people who lsiten different types of music react in different ways on different channels, and you cannot really measure the success properly.
When it comes to the prediction property of the HSS, I believe that this is a good idea, and I am really curious if someone has already done that. Still I do not know if that can happen.
I find the topic really interesting, and I really want to research examples.
Very interesting post which led me to search some more information on the outlook of a potential “commercially succesful” song. Whilst doing so I found a blog post examining a tool named the “Infinite Jukebox”. This tool analyzes the beat pattern of a song by using different colors and drawing connections between similar beats. Since I wasn´t able to copy pictures in the comment box, hereafter follows the link to the blog post with the beat patterns for many well known pop songs (i.e. “Rolling in the Deep”, “Baby”, “Tik Tok” and more).
Interestingly, all of those “commercially successfull” songs follow a similar pattern, caracterized by a long intro and first verse, a shorter second and third verse, a chorus with two repetitions and an outro (please have a look here at the link I posted above for the visualization of the typical pattern).
Relating this to the blog post of Euclid Alexis Haralambidis, I think HSS will definitely be a part of the music industry also in the future as it allows to gain insights into the framework a song should have in order to be potentially “commercially successful”. However, the framework alone doesn´t guarantee the song will actually be successful since there are many more elements that come into play before a new song actually reaches the top charts. To just name a few: marketing, reputation of the artist, moment of release and not to forget the “human” elements of the song consisting of the lyrics and the voice of the artist.
In conclusion, I like to think of HSS as a toolbox allowing us to determine the framework of a successful song (like for example an essay should have an introduction, a main body and a conclusion). But still there are more factors we need to consider before we can predict a songs success with certainty.