Genre recognition has become the holy grail of music information retrieval. What concerns me, before we worry about machine recognition of musical genre, is whether people can agree at all on what genre means, and what the various genres are. Wikipedia, Echonest, and many other sites (some now defunct) have put forth excellent information on various musical genres and their relationships to one another. My critique of the genre discussions I have encountered to date falls into two categories. One is the (necessarily, and not surprisingly) culturally narrow perspective of most work on musical genres. The other is the role non-aural, non-audio features play in the determination of genre. (These can be metadata, like release dates, or even more [sub]culturally determined information such as the clothing style of the artists.)
Let’s take the problem of narrowly culturally situated efforts first. There have been a variety of impressive resources on the Internet about the sub-sub-sub-genres of electronic music and of extreme metal. There is a wonderful degree of detail provided in these Web resources. However, the effort put into very subtle distinctions among “northern-based” musics (anything we typically understand as pop music, plus the folk and court musics of northern Europe and North America**) is rarely, if ever, matched by the knowledge available, perhaps, in English, on musics from other countries. We typically find some half a dozen genres listed for Brazil, Mexico, Japan, or Cuba, and far fewer for China, Turkey, Belize, Honduras, or Mali. This is a typical case of out-group bias, which is easy to understand; all people are subject to out-group bias. The importance of understanding biases lies in the effort to move beyond them. Are the differences among Xote, Brukdown, Özgün Müzik, and Guarapachangeo less significant than the differences between Goa Trance and Happy Hardcore, or Grindcore and Power Violence?*** Of course not, but who can know every little detail about the impossibly rich musical landscape of every culture? (That’s why we need multi-cultural teams to work on genre recognition and classification.)
The other issue is one I am only aware of in terms of “northern” (western) popular forms of music, and it is the issue of image-based, fashion-based, temporal, and geographical genre attribution. In many cases, the clothes worn by rock and pop artists seem to determine their musical genre more than the sounds created and organized into musical works by those artists. For example, Billy Idol and Avril Lavigne are thought of as Punk Rock artists. Yet, and even without appealing to DIY ethics and political content, we can tell from the aural experience that these artists make (or have made) something sufficiently aurally distant from the music of CRASS, pragVEC, Buzzcocks, X-Ray Spex, or BAD RELIGION, and that theirs are genres well removed from Punk Rock. (The artists listed do not all sound the same, but they share the elements of disaffected vocals, a lack of polish, and an overall dark despair with one another and with bands as far removed from them as Joy Division, The Paper Chase, Depeche Mode, and Sleater-Kinney, all of which have more sonic elements in common than they do with Idol or Lavigne.)
What makes the problem further difficult is that genre names are rarely descriptive, and all too often temporally and geographically limiting. Consider the genres NWOBHM (New Wave of British Heavy Metal), New Wave, Nü (new) Metal, Grunge, and Old-School Hip Hop.
Quite apart from the problem that “New Wave” actually has at least three different meanings, it is sonically possible (and common) for an artist making music thirty years after the end of the era attributed to one of these genres to make music with the same structure, affectation, instruments, sounds, and production. Which should we consider in determining genre: the year of release or the way the music sounds? New-millenial bands like Titanium Black and The Haunted, and even punk-rockers like Saviours, often play a flavor of Metal that sounds just like NWOBHM, but we are not supposed to call them that if they are from a different time, and especially, a different place. Likewise, ACCEPT and SCORPIONS (from Germany) sometimes played the same type of music, stylistically speaking, as Judas Priest, DIO, and IRON MAIDEN, but since they’re not British, we cannot refer to their music as NWOBHM. Or can we? Is it not the sounds and how they are organized that matters in determining music? (I think so.) Can anyone really tell, in a blinded listening test, whether a rhythm guitarist is German or British?
The Union Underground was a Metal band that had some success during the Nü Metal years. They had the look and the album art to be part of that era and that genre. However, listening to their music in 2008, I could not help but notice that the singing style really had little to do with Nü Metal, and quite a lot to do with Grunge, which was declared over by that time. As far as I can tell, no one talked of TUU as a late Grunge band.
An interesting pair that got me thinking further about image- and time-based genre attribution are Corrosion of Conformity and VOIVOD. Originally starting out in very disparate genres, in the farthest reaches of Hardcore Punk and Prog Metal, these two diverged in their music until their releases of the albums ‘KATORZ’ (by VOIVOD) and ‘CORROSiON OF CONFORMITY’ (by CoC, of course) in the late aughts. I find it nearly impossible to tell these two albums apart stylistically (though each is quite distant from the bands’ earlier output). When I saw CoC perform at Dante’s in Portland, they presented a marvellous synthesis of Prog agility and Punk attitude. (These two were not meant to go together, but it’s happening more and more.) Meanwhile, VOIVOD apparently drifted further and further into Punk Rock, and lost most of their Prog intricacies. Yet, if I were to stick to “what we know those bands to be,” I would be forced to attach opposite labels to songs from those two albums, which, even when I’m looking right at the readout on my display and know what I’m hearing, sound the same to me.
I mentioned Old-School Hip Hop above as well. Every now and then, you hear a new song, and it has that early, innocent flow we associate with everyone from The Jungle Brothers to MC Hammer. It’s old-school in that the time extents of rhythmic phrases in the vocals and the time extent of semantic phrases in the lyrics delivered by the same vocals coincide****.
Yet, maybe it was released in 2013. Yet, De La Soul was putting out music in 1989 that did not sound old-school; it was like what was going to happen ten years later. (I feel the same way about fu-schnickens’ 1992 album.) Some of the music in those old-school days was well ahead of its time, and some music that gets released even today brings back the old-school style. It’s the sound that counts, not the metadata.
There are many more examples, and perhaps better ones that I will add as I think of them, or hear them, but for now, I will conclude that, 1) genre studies and genre R&D need multi-cultural teams so that the level of attention to detail that is possible for Deep Psytrance vs. Gabber vs. New Romantic vs. New Wave will also be possible for ‘Bulgarian Rock’, ‘Hungarian Rock’, ‘Russian Pop’, and ‘Turkish Pop’. Sure, I’m glad someone in America even cares enough to put those on the map, but given the several hundred varieties of Electronica, Metal, and Hip Hop each, can we really believe there is only one variety of ‘Russian Pop’? (I know for a fact there are quite a few styles and genres within Turkish Pop.*****)
* Yet, no matter how much detail each scholar, researcher, developer, or enthusiast goes into, it is likely to prove insufficient for the afficionado of that sub-sub-genre. The Echonest blog (at http://blog.echonest.com/post/52385283599/how-we-understand-music-genres) recently included the following comment: “. . . somebody, somewhere might care about (e.g. “gothic metal” vs. “symphonic metal” vs. “gothic symphonic metal”).” Yes, somebody right here not only does, finds those to be rather obvious and relevant distinctions, which are further complicated by Nightwish’s recent experiments combining symphonic, power, and folk metal.
** Why I use the term “northern” rather than “western” will be the topic of another post.
*** Yes, these are real genre names. To a connoisseur who lives Grindcore, Power Violence may sound completely different, while to an outsider neither would be distinguishable from the earliest ’80s Thrash.
****As Hip Hop matured, it became less and less common for the sentence and its rhythmic phrase to start and end together. This seems to be a result of the recognition that rhythm allows one to rhyme with any syllable in a word, not just the last syllable. So, an MC who wants to rhyme, for instance ‘crime’ with ‘time’ does not need each sentence to end with one of those words; s/he can rhyme ‘crime’ with ‘time’ in a sentence that might go “It was that time [break here] I went off to the east coast” where ‘time’, due to rhythmic phrasing, took care of the rhyme, and the rest of the sentence could still be uttered. In much old-school Hip Hop, the semantic phrases had to end at the same rhythmic stopping point.
*****As for style versus genre, let me try a quick explanation. A guitarist can play the blues in a Be-Bop context, a Psych context, or a Funk context (to name a few), and an MC/toaster can rap on a Hip Hop song, a Reggaeton, a Rock or Metal song, or even in a piece of modern “classical” music. Similarly, a drummer can play funk, swing, or shuffle in a Jazz band, Rock band, Pop group, or an experimental combo. The elements these musicians bring in are styles (blues, rapping, funk, swing, shuffle), while the complete package of the musical experience will likely fall into a genre or subgenre, like Electro Swing, Funk Metal, Be-Bop, or Chorinho.