Music Criticism as a Creative Tool

Paseando por Internet me encontre con este link muy curioso: www.pitchformula.com, que se trata basicamente de un proyecto universitario muy curioso.
Básicamente,un estudiando se puso como tesis lo siguiente: armar música basandose en el gusto músical que se podria sacar de reviews de discos de un sitio en particular, en este caso, la infamous revista Pitchfork Media. Armando un complicado sistema y bajandose la base de datos de reviews de discos, basandose en el puntaje de cada disco, y las palabras que se usan, y en que contexto, empezo a extraer palabras y conceptos que eran "buenos" (causaban una review positiva) o "malos" (causaban una review negativa).

Another interesting thing is the presence of processing/effects words (ambience, effects, processing, and reverb); apparently, Pitchfork reviewers are very fond of recordings that have artificial effects on the instruments. There are also several words which point in more “experimental” directions (noise, drone, feedback, and silence). “Noise” is more than half the score of “melody”, and “drone” isn't far behind, so the Pitchfork critics definitely want to hear some experimental elements in their pop songs.

Todo eso es ya de por si bastante colgado y en general las conclusiones sobre los gustos son bastante predecibles (ya que vamos, cualquiera que entro 3 veces a la pitchfork sabe mas o menos por que van las cosas que le gustan y las que no), aunque tambien hay algunas otras conclusiones que saco. Obviamente, todo muy vago, porque es basicamente información estadistica.

I find myself wondering why a critic would be more likely to mention ubiquitous instrument words like “guitar” when they're talking about a record that they like, and then not mention those words when talking about a record that they don't like. This is likely connected to the phenomenon that the Pitchfork critics are 3.8 times more likely to use a meaningless value judgment word when they're describing something they don't like. The critics are much more likely to describe the music itself when they like it; if they don't like the music, they aren't nearly as likely to go into detail about it in the review.

When the Pitchfork reviewers write about a record they don't like, they're more likely to mention two things:

1. words referencing consumerism and business, and
2. words insulting the intelligence of the musicians or listeners.

The former expresses an anti-capitalist slant, and the 9 words that I identified as part of this group have a significant wordscore when added together (-4661.02; the word “sell” alone scores -1432.56). These anti-capitalist words mostly refer to advertising, marketing, and selling music.

Tambien aprovecha para darle un palo al sitio...:

On a related note, the small number of jazz records that are reviewed on Pitchfork have a much higher average rating than the other records. The artificially high ratings of records in a certain genre points to a fear of bashing these records, perhaps due to a lack of skill when criticizing these records. Alternately, maybe Pitchfork only chooses to review known-good jazz records, which makes sense since they aren't a jazz publication and wouldn't feel obliged to review all the new material in that genre.

Luego de todo este cuelgue recontra nerd paso a la otra parte del proceso, que fue básicamente componer 2 temas basandose en los resultados, algo asi de un sistema para hacer un disco pitchfork-complaint. Afortunadamente el chico es bastante habilidoso con la música y los resultados son interesantes. Y si, suena a un disco que podria ser analizado en la ya mencionada revista. Me parece igual super interesante el hecho de usar lineas tan estrictas para armar música. Siempre me parecio super interesante el hecho de usar limitaciones (en este caso tan poco convencionales, como el supuesto grupo musical de un sitio web) para componer.