My Not-So-Secret Software Project

I've been cagey about what I've been working on. Maybe what I've felt is akin to that superstition of some writers: that talking about what you're going to write soaks up creative juices that should be oiling the machinery of production. But I've told too many friends now, and besides, I'm stuck for a blog entry between my twice-daily Uberman On-Ramp entries. I don't want my blog to become All About My Weirdo Sleep Schedule. It's not as if you can't work toward uberman and still be an interesting person. Consider PureDoxyK (the mother of the movement, such as it is), or self-made self-help guru-blogger Steve Pavlina.

My Project X? Here's what it is: a Japanese-to-English machine translation tool that even professional translators wouldn't scorn. [He waits for the laughter to die down, scowling, thinking "You laugh now, but when I'm rich ...."] I started work on it in September, let it lapse in December and January, but have since been working on it almost every day since early February.

Here's the problem: much of the job of being a translator is already fairly mechanical. Of course, if you're translating Japanese that's semantically nuanced and culture-bound (novels, poetry, ad copy, movie subtitles, manga, anime), you'd better be full-spectrum human, and quite fluent and literate in both languages. But for anything technical or legal, it might even be an advantage to have a touch of Aspberger's Syndrome.

I've mostly translated patents, which are both technical and legal. Japanese patents run very much against the grain of vernacular Japanese, which is, in daily use, a notoriously vague and allusive tongue. Patents are grindingly explicit. In fact, the only distinctly human thing I've found in patents is the phenomenon of invention itself. (And not as often as I'd like; see below.)

I know I've got a good patent when I'm confused while reading it, only to have the proverbial lightbulb light up in the foggy comic-book dialogue-cloud above my head: "I see! The reason I was having a hard time understanding it is that this is a real patent: an explanation of a new and counter-intuitive technological idea."

Then it's fun. Then I'm rooting for the patent filers, because I feel they deserve their claim and any royalties accruable, for Bringing Fire to Man. I'm grateful for my bit part in the relay race -- I get the torch, I carry it my little part of the way, I pass the torch, but before I pass it, I sit huddled in my dark cubby after midnight, before my glowing screen, and blow on it and watch it flare a little. At those moments I feel more a part of the great human story than I ever would have felt as an altar boy on his first day, or if I had joined the Cub Scouts. (Two things I never did and never wanted to do -- I'm not a Joiner.) If only all patents -- but especially software patents -- were actually for real inventions. But that technological Valhalla must be pulsing and throbbing away in some other part of the galaxy. Here, we chip flint, we grovel before those who claim a slightly new style of crafting arrowheads, we pick lice out of their hair, we polish their arrowheads and crow over the hypocritcally, and hope they toss us the leftovers of their mammoth marrow pudding.

At its worst, patent translation is a special circle of hell: the patent text and claims are so long, the language so convoluted, the idea so evanescent (or non-existent -- many Japanese patents go on at great length about the obvious and nothing more), and the translator (that would be me) so tired from translating so much nothing, that you just start obsessively reviewing, wondering if the idea is so brilliant but so counter-intuitive that you just can't get it. This is especially bad when you've already blown one of those soul-crushing deadlines. Which I never do. Ever. (Well, sometimes.)

OK, so that's the problem: coming up with machinery that helps get the soul-crushers out of the way, and also helps you with the good stuff. Maybe it reduces the amount you have to type -- some translators have resorted to dictation to avoid further tendinitis damage. Maybe it helps you break vocabulary blocks. Maybe it helps you juggle parts of long sentences without doing damage to the syntax. Maybe it helps you automate sweeping, complex revisions. Maybe it makes sure you don't miss anything. I'd eventually like something that can figure out things about the diagrams, but that might be far too ambitious ona project that's already very ambitious.

There are several problems that I hope this project solves for me. One is that I'm in a sabbatical from translation, hoping to improve my skills in Japanese before I take on work again. Another is that my dream life is to make my living while traveling. And any software that takes the edge off, and helps me meet my deadlines while I'm also contending with the stress and logistical complications of life on the road is all to the good. Also, software was about the only thing I ever got professionally very good at. (I just fake being a hotelier -- so far, it's working.) Software is still a career option. But only if I keep my skills sharp. Alas, they've gotten rusty.

Finally, there's just what I most loved about writing software as a craft: messing around with complex data structures and algorithms. I'm sure I'll get my fill of that with this project. I like translating patents, but mainly because I like invention. I want to contribute inventions of my own, even if it's in a field (software engineering) where invention seems to be slowly waning, where doing anything new is getting harder every year.

I think I'm two years away from a production tool that others could use. Maybe one year away from having something I can use in my translation work, while continuously improving it. I've considered making it open source, but I'll cross that bridge when I come to it. Right now I'm bogged down in just getting the thing to parse basic Japanese. I'm giving myself three more months to finish that. Then what? Yes: parsing intermediate Japanese.