Most of the work I’ve done from the statistical side so far has been based around the simple idea that looking at lyrics data can help get some sort of insight into the music of any given rapper. The artists’ work, through this lens, is the only thing that matters. This way, any conclusions made about a rapper is inferred from the actual lyrics and nothing else.
Obviously this is a severely limited type of analysis. Certain guys can market themselves as lyrical when really, empirically, nothing they do is that impressive. Conversely, certain rappers considered ‘simple’ or ‘garbage’ for their lyrical content are in fact empirically complex. The perception of the fans in this way is just as important as the lyrical content.
Gauging artist perception from the fan/listener P.O.V. is an awkward task. How do you get good, representative data? How do you express this data in a way that makes sense? Luckily, there’s already a good foundation for this type of stuff. Sentimental Analysis is a field in natural language processing that tries to put objective measures to the attitude of the commenter. This way, a review can be analyzed statistically to determine the feelings towards the subject on some sort of scale; negative or positive, for example.
Reviews, especially by music critics, aren’t all that great as a resource for a few reasons. The simplest one is that there just isn’t a lot of great rap writing to mine. Most rap websites serve as content farms and it’s far more efficient, from a web economic standpoint, to post 10 new songs a day than it is to write in-depth reviews that nobody will read or care about. The more important factor though is that a ‘star-rating’ sentimental analysis might miss the point. It’s just way less interesting to know that a troll-wave Riff Raff is a ‘2-star’ rapper than it is to understand what his fanbase consists of.
A ‘star-rating’ system for a movie might affect who watches it but it doesn’t tell us much about why the people who do like it like it or where they’re coming from when they do. Generally speaking, rap fanbases seem fractured and exclusive. A’ street’ rapper might be called dumb by ‘backpackers’ and a ‘backpacker’ might be called a lame nerd by everyone else. To me, the backpacker/street divide is one that deserves some empirical analysis.
To do this, I mined Youtube comments from about 500 different rappers. I attached certain keywords with each side of the ‘backpacker’/’street’ stuff. Words like “lyricism”, “real hip hop”,”nigga”, “fucks wit” were strong indicators. For each artist, I scanned 10 relevant videos with a high volume of comments. Then put a score to it.
Some obvious problems arise from this methodology:
- Are the videos representative of the artists’ total body of work?
- Are Youtube commenters representative of an artists’ listenership?
- Are the keywords themselves good indicators?
Clearly a good study would try to account for these problems and the other 100 problems that arise from this but fuck it.
So, first, we have two distinct artist types we want to separate. Pure ‘backpacker’ rappers and pure ‘street’ rappers. Table 1 shows each categories’ top 20 artists. It passes the initial eye test. There doesn’t seem to be any artist that definitely doesn’t belong. Also, if that list of 40 rappers was given to a rap listener and they were told to put these artists into two distinct categories, it would probably resemble the lists below.
|4||C-Murder||Lords of the Underground|
|5||Koopsta Knicca||Rah Digga|
|6||Suga Free||Chubb Rock|
|7||Boyz N Da Hood||One.Be.Lo|
|8||Lil Boosie||Talib Kweli|
|10||Yo Gotti||Grand Puba|
|11||C-Bo||Kool Moe Dee|
|12||Yung Ro||Das EFX|
|13||Soulja Slim||Crucial Conflict|
|15||Guerilla Maab||A Tribe Called Quest|
|16||Mac Dre||The D.O.C.|
|17||Lil’ Keke||Poor Righteous Teachers|
|19||Dayton Family||Organized Konfusion|
|20||Kane & Abel||Jeru the Damaja|
As a quick and dirty tool to seperate styles, this works pretty well for extreme cases it seems. It becomes apparent that Youtube comments have a direct relation to the type of music and that certain keywords serve as good tests to learn something about an artists’ fanbase. I’m surer there are modular tools that could be made to look at other genres in the same way but that’s dependent on those genres having some kind of binary divide in fans, I think.
At this point, without really listening to an artist you could learn something. For example, somebody online suggested that this group Clear Soul Forces was the best new group he’s heard in years. It seems derivative of NY Premo scratching sneakerhead soft stuff but I wasn’t sure. Clearly they’re not ‘game changers’ by any means. They scored safely into the ‘backpacker’ range. ALGORITHMIC HATE.
If you want to play around with some of the data, use the link below: