Editions, Tweaks, and User Preferences
Column Editor: Michael P. Pelikan (Penn State) <[email protected]>
I’ve made comments before in this space about problems that continue to plague eBook projects that begin with out-of-copyright print sources. Optical Character Recognition (OCR) has improved hugely over the past ten or fifteen years, but achieving the last incremental improvements that would bring it close to practical perfection has proven difficult. Even if achievable, near-perfect OCR would do nothing to address the backlog we’ve accumulated of poor OCR’d texts, many of which, as mentioned, are out of copyright.
This means there’s not a lot of financial incentive to promote investment in retrospectively repairing past results of flawed OCR projects. This came up for me again recently whilst reading, for only the second time in my life, the Personal Memoirs of Ulysses S. Grant.
My first encounter with this material was through Project Gutenberg. It came in the form of a pure ASCII text file. It had line endings and carriage returns, but nothing more exotic than that. The file itself was not the product of OCR. Instead, it was typed by true enthusiasts: candidates for sainthood who felt strongly enough about a particular book to take on the task of transcribing as an entire work from printed page into keystrokes, for the good of the World.
The quality of transcription of many such works was variable, but improved over time. This was not in small measure because other folks came along and began to make corrections to the hand-built editions, in a way somewhat similar to how a wiki article can be improved over time. Better, in some ways, because there were fewer matters relying upon subjective interpretation, at least in the case of same-language transcriptions — either it was correct or not.
I don’t really understand, if a human-generated, even curated, transcription exists, why the builders and publishers of e-texts don’t take advantage of them. Why start from scratch and apply machine-driven OCR to printed text if there’s already a transcription? Many, perhaps most, such transcriptions are freely available and could be used — it would cost only attribution and recognition of the source, something I’d perhaps wrongly assume that even the most craven, financially motivated republishers of old works could bring themselves to do.
Instead, now, a dozen or more years after admiring the transcription of General Grant’s memoirs, and hoisting a coffee cup in toast to the unknown person or persons who made it possible for me to enjoy the work, I’m confronted with obvious, characteristic OCR errors in a recent eBook edition. Grumble.
But this shouldn’t be the end of the story! Have you noticed that Kovid Goyal’s Calibre (http://caliber-ebook.com) permits the editing of an eBook file? Regular readers of “Antidisambiguation” (at least, those who would admit to it) will recognize my shout-out to this extraordinary open source software package. If you use an eBook reader, I mean, at all, you owe it to yourself to have a copy of Calibre installed somewhere.
All right, but say I use Calibre to fix an obvious OCR botch in an out-of-copyright work like Grant’s — what then? Well, I’d have to sync the repaired file to the several eBook readers I maintain, as well as the file servers I keep at home for purposes of redundant backup. Ever looked into NAS RAID devices? These are a faintly miraculous technology, once accessible only among the corporate or the hopelessly geeky — now available to all! I presently employ three of these boxes on my home network, each containing two hard disk drives configured to mirror each other. Whilst they quiet the mind, they also exact a bit of overhead in terms of file management — but good file management will always entail a blend of good decisions and good practices.
The idea of applying corrective measures to an eBook differs only in degree from things we already do. Those controls on your audio devices labeled Bass and Treble? Those have been collectively referred to in the past not merely as Tone controls, but Equalization controls. The concept behind audio equalization is corrective. Recognizing that different listening environments have differing acoustic characteristics, as do the many and various transducers in use, thoughtful manufacturers of audio gear provided audio controls permitting one to tailor the frequency response of one’s audio gear to compensate. If your rugs and curtains absorb high frequencies resulting in, say, a six dB roll-off at 10 kilohertz, you can boost the response of your system at 10 kilohertz by six dB to “equalize” it.
Of course, many folks don’t use these controls to equalize anything but, in fact, to de-equalize, indeed, to change the frequency response of their audio systems simply to suit their preferences. Those worthies cruising slowly down the street in the low car with dark windows and after-market muffler, whose audio system’s subwoofer can be heard two blocks away, sending ripples through puddles like Crichton’s T-Rex, melting their tympanic membranes — they’re merely applying user preferences.
This appetite to configure, to tweak, to personalize, must cause despair, or at least shrugs, among the engineers and producers who struggle to achieve a particular sound in a produced recording. The thoughtless destruction of producer’s and artist’s wishes has been going on for a long time. Ever been in a discount store and heard one channel of a stereo recording in housewares and the other in lawn and garden? I recall a story my brother told of the fourth and last time he went to Stanley Kubrick’s “2001, A Space Oddessy” — it was in 1969 at a drive-in theater in Indiana. It was raining heavily. You could just make out the screen through the fogged windows. The little metal speaker box hanging in one side window was struggling to handle “Also Sprach Zarathustra” with little success. Poor little thing…
I’ve long wished for there to be released the audio version of critical editions of recorded classics. As a darn-near-life-long multitrack audio production guy, there’s nothing I’d like more than to get my hands on a multitrack version of particular classic recordings. As soon as the Beatles got past “Beatles ’65” they were increasingly taking advantage of technical possibilities afforded them by their studio, and opened by the skills of George Martin. Hendrix’s early recordings were very simple. In the space of a few hundred days these artists were taking their music places few had gone before, and they were layering sound upon sound to do so. It was the audio equivalent of photo or motion picture compositing, placing elements of differing origin into seamless proximity with each other.
With a multitrack edition of these recordings, one could separate the original signals, listen to each individually, and gain a better understanding and appreciation for how the producer and the artist achieved such phenomenal results. Of course, it would require that a multi-channel mixer be part of the signal chain — but who wouldn’t want that? And if a particular sound always seemed buried to you, you could bring it out in the mix! Conductors do this when they interpret a score in front of them, shaping the statement and balance of each of the parts of the score through guidance provided to the orchestra. Really, a musical score is a multitrack representation. So its counterpart in recorded music — that’s all I’m asking for…
Blu-Ray and DVD editions of motion pictures often offer options in playback to include or exclude deleted scenes, to change language settings, etc. I’ve seen the occasional book, usually a children’s book, that feature branching in the storyline, permitting exploration of alternate plotlines based upon decisions as you go.
I know it will probably not happen in my life time. Works of interpretation are works themselves — that’s probably part of the reason why such a great idea won’t easily come about. Royalties and Intellectual Property issues involving derivative works get complicated. But I’d be happy to sign a license attesting that I would not release a remix of Sergeant Pepper or Electric Ladyland — I would only take bits of them apart to see how they work. This isn’t too different from standing in front of an artist’s masterpiece in a museum with a sketchpad, working with charcoal and paper to understand what’s going on in the painting or sculpture.
There are some promising prospects enabled by digital audio analysis. Some of the same algorithms that achieve noise removal through example (sample the offending waveform, then look for it in compound waveforms and separate it out, leaving a clean signal) can be used to “de-mix” a mixdown. It might be feasible before long to divide a favorite recording back into separate tracks.
If you’re interested, there’s an intriguing PhD dissertation at Stanford’s Center for Computer Research in Music and Acoustics entitled “Interactive Sound Source Separation” by Nicholas J. Bryan. The dissertation is licensed under a Creative Commons Attribution-Noncommercial 3.0 United State License. Google that title to find the pdf. Outstanding work.