I am interested in doing some natural language processing on copyrighted song lyrics. I saw a related project here, http://rappers.mdaniels.com.s3-website-us-east-1.amazonaws.com/ who used rap.genius.com to get lyrics. There is a related data product available for sale as a poster (with no copyright/other licenses mentioned in the footnotes).
(1) Does anybody know if this is a legal use? Would it be equally valid if it were copyrighted novels?
It appears that song lyrics are copyrighted https://writers.stackexchange.com/questions/6797/are-music-lyrics-copyrighted but none of the documentation I found talked about data products made from them. Obviously I would need to house a corpora of copyrighted material to produce aggregate information, and I want to know what the data use rules are for lyrics not in the public domain.
(2) Does anybody know what legal rules you are subject to by processing copyrighted material in bulk? This seems to be the parent question? (Any comments relating this question to google's book scanning project are welcome)
A legal basis would be nice. I found a discussion of UK rules, http://copyrightuser.org/topics/text-and-data-mining/ but I'm interested in the US rules. It looks like this document from the US copyright office has some answers, but I haven't gone through it yet: http://copyright.gov/docs/massdigitization/USCOMassDigitization_October2011.pdf Anybody have a summary of that document?