Help teach my computer to follow a chart
August 12, 2010 8:50 PM Subscribe
Are there any machine readable music "fake books" available, free or for sale?
I'd like to do a statistical analysis of chord progressions and I need a large chunk of data to work with. An electronic Fake Book or Real Book seems like a good source of hundreds of songs in consistent notation. All I can find are PDF distributions though, which can't be easily parsed. Are there any electronic versions available in plain text or MIDI? Or, say, XML?
Keys to making this work will be consistency and accuracy of data and ideally the simple lyrics/melody/chords format. MIDI collections gleaned off various sites aren't much use due to widely varying quality. Emphasis is on jazz standards, although large collections encompassing pop or a variety of styles would also be nice. The more songs the better!
I'd like to do a statistical analysis of chord progressions and I need a large chunk of data to work with. An electronic Fake Book or Real Book seems like a good source of hundreds of songs in consistent notation. All I can find are PDF distributions though, which can't be easily parsed. Are there any electronic versions available in plain text or MIDI? Or, say, XML?
Keys to making this work will be consistency and accuracy of data and ideally the simple lyrics/melody/chords format. MIDI collections gleaned off various sites aren't much use due to widely varying quality. Emphasis is on jazz standards, although large collections encompassing pop or a variety of styles would also be nice. The more songs the better!
Interesting idea -- there's probably a CD-ROM with .pdfs of all sorts of real books floating somewhere near you (ask the tech-savvy musicians). You could conceivably run OCR over them and get some good data -- http://code.google.com/p/pytesser/ might get you started. Some savvy searching might turn up lead sheets in MusicXML too.
posted by gunslingingbird at 9:46 PM on August 12, 2010
posted by gunslingingbird at 9:46 PM on August 12, 2010
You could try using some of the songs on the Ultimate Guitar Archive as a source (link is to the top 100 in popularity).
If all you want is the chord progressions (without any reference to their relationship with lyrics or to their timing) then this might work for you - in most cases your raw data is just sitting there in alternate lines of the text.
If you want to start with printed music (or PDFs of printed music ) then PhotoScore Ultimate might be useful.
posted by rongorongo at 2:43 AM on August 13, 2010 [1 favorite]
If all you want is the chord progressions (without any reference to their relationship with lyrics or to their timing) then this might work for you - in most cases your raw data is just sitting there in alternate lines of the text.
If you want to start with printed music (or PDFs of printed music ) then PhotoScore Ultimate might be useful.
posted by rongorongo at 2:43 AM on August 13, 2010 [1 favorite]
Yeah, I was going to suggest the guitar tablature sites as well. They are definitely more oriented towards pop/rock/metal but there is still a lot of diversity and they are all plain-text files (with the exception of the stupid people that upload them in some proprietary Guitar Pro bullshit, grrrrrr) which makes parsing a lot easier. There's still a lot of variation so you'd probably spend a fair amount of time automating the extraction process but it has to be easier than OCRing the hand-scrobbled fake books.
posted by Rhomboid at 8:18 AM on August 13, 2010
posted by Rhomboid at 8:18 AM on August 13, 2010
The Real Book jazz standards have long been available in Band In A Box format online. Since the Real Book isn't licensed, neither are the BIAB files of them. If a simple search doesn't find what you're looking for, look for a torrent.
You'll have problems with quality, and with interpretations of songs, though. A decent percentage of user inputted Real Book charts are creative interpretations at best.
You'll also have problems generating useful data if you don't standardize your sourcing. In example, the Real Book conflicts with the New Real Book. The New Real Book has the correct chords as approved by the composer, since it's a licensed work, while the Real Book has the chords that made sense to the Berklee students who did the transcription from a particular performance (unlicensed).
In a typical jazz tune this may be as simple as the difference between tritune substitutions in the RB and 2-5-1 in the NRB. But in latin standards Bonfi, Jobim, etc., you'll see entirely different interpretations of the chords. You'll at least want to listen through your database and decide which tunes match your ears' criteria for "correct" or "good" chord changes.
In short:
torrent real book
posted by lothar at 8:43 AM on August 13, 2010
You'll have problems with quality, and with interpretations of songs, though. A decent percentage of user inputted Real Book charts are creative interpretations at best.
You'll also have problems generating useful data if you don't standardize your sourcing. In example, the Real Book conflicts with the New Real Book. The New Real Book has the correct chords as approved by the composer, since it's a licensed work, while the Real Book has the chords that made sense to the Berklee students who did the transcription from a particular performance (unlicensed).
In a typical jazz tune this may be as simple as the difference between tritune substitutions in the RB and 2-5-1 in the NRB. But in latin standards Bonfi, Jobim, etc., you'll see entirely different interpretations of the chords. You'll at least want to listen through your database and decide which tunes match your ears' criteria for "correct" or "good" chord changes.
In short:
torrent real book
posted by lothar at 8:43 AM on August 13, 2010
This thread is closed to new comments.
As an upside, you may start to notice some patterns yourself just inputting the chords.
posted by fantasticninety at 9:12 PM on August 12, 2010