The general feeling among researchers is that translators will continue to play a central role in production of the high quality translation well into the future. They will also inevitably contribute to the fine-tuning and repairing of MT output as post-editors through the feedback loops that are vital to optimizing MT systems. The gradual build up of postedited texts will then turn into a huge body of potentially decisive training data for MT systems.

There will naturally be more research into ways in which this symbiotic relationship can be optimized within the various types of workflows, with improved toolsets for post-editors. But it seems unlikely that there will be anything more than incremental advances in performance for the industry as a whole. We can expect forward-looking technical translators to adopt new power tools emerging from such research to stay competitive.

[snip]

As to the old fantasy of the perfect artificial translator, the hypothesis on the table is that a system capable of systematically aping (or even surpassing) a human translator will need to draw on ‘world models’ – real-world knowledge – to overcome the critical quality bottleneck. But it has so far proved impossible to program a machine to understand the semantic intentionality of a text.

Nice to see some writing on this issue that doesn’t shy away from the fact that as good as the computers are likely to get, they aren’t going to approach the depth of human mastery and replace us “any day now.”

Chop into pieces; cry.

This final editing stage can be tough. When I handle it on these texts, I might be dealing with the output of three or four translators. It’s a chore to go through all their contributions, making sure the serial commas are struck out (if the client doesn’t like them), adjusting spellings to US or UK English as needed, and making sure we don’t call a “law” an “act” five pages later. And the clock is ticking.

It’s possible to turn out a quality translation in these conditions because I’m dealing with (a) talented, experienced translators, (b) who have all worked on similar documents in the past, and (c) all had access to the entire text as they worked, so they could confirm how their chunk fit into the whole or note that the Ministry of Economy, Trade, and Industry would be spelled out in full on page 2 so they could use “METI” on page 14. But there are a growing number of companies selling crowdsourced solutions that can’t afford (a), can’t guarantee (b), and actively reject (c) as part of their work process.

I’m afraid I forgot who brought this to our attention on JAT List, but here it is: from the New York Times, “At Microtask and CloudCrowd, Assembly Lines Go Online.” Innovative businesspeople finding ways to innovate every last drop of quality right out of a translation by breaking it into as many parts as possible and giving nobody a look at the whole text:

CloudCrowd, based in San Francisco, also offers to distribute clients’ work online. Like Microtask, it has found ways to break work into thin slices.

“Rather than crowdsourcing, we call what we do widesourcing,” says Mark Chatow, the company’s vice president for marketing. “We take tasks like translation that used to be done by a single specialist and break them into pieces so a wide range of people can handle different parts of the work.”

CloudCrowd uses machine translation software to make a first pass. Then it sends out individual pages of the machine’s translation to garble hunters, who look for sentences containing a nonsensical sequence. A translator with native language fluency is needed only for the sentences tagged by the garble hunter. An editor, without foreign language expertise, then polishes the prose, but possesses only a single page, not a chapter or the entire work.

CloudCrowd exclusively uses Facebook members who come to it for assignments; it says it has 50,000 workers in its crowd. Traditional translation costs about 20 to 25 cents a word, Mr. Chatow, says, but “we’re doing it for 6.7 cents a word.” He says translators make an average of $15 an hour and garble hunters around $7 an hour.

And then comes the money quote:

Mr. Miettinen of Microtask says, “Pure monetary compensation is a 20th-century concept.” He envisions tapping the talents of game designers who would render clickwork fun, what he calls “game-ification.” If successful, it could minimize complaints about pitiful pay or soul-draining boredom.”

Now there may indeed be a major market for this stuff—texts for which clients don’t have a budget for real translation, but which need to be a little bit more readable than Google’s MT gibberish. But it remains to be seen whether 6.7 cents per word is the magic point at which the total lack of consistency and polish across your hundred-translator document can be overlooked in favor of the vaguely passable grammar in each sentence.

It also remains to be seen whether Mr. Miettinen remains mired in that twentieth-century concept of paying himself a handsome salary from his pool of venture capital.

The AI crowd, too, was pleased with the result and the attention, but dismayed by the fact that Deep Blue was hardly what their predecessors had imagined decades earlier when they dreamed of creating a machine to defeat the world chess champion. Instead of a computer that thought and played chess like a human, with human creativity and intuition, they got one that played like a machine, systematically evaluating 200 million possible moves on the chess board per second and winning with brute number-crunching force.”

A nice encapsulation of the way machine translation is moving forward—through raw computing firepower and improved statistical approaches, not through development of computers that can truly do the human translator’s job in human ways. (Yet? The jury’s still out here. There are many orders of magnitude more possible moves in language than there are on a chess board.)

Now Kotaku is a tech blog, so of course its writers will focus on the tech aspect of the story being told—here the iPad comments from a well-known animator. To its credit, though, it includes a solid paragraph presenting some of the additional nuance provided in the Japanese article it references:

He might seem like an eccentric technophobe, but he is coming from an “All I need are pencil and paper” point-of-view. That might be all he needs. He’s Hayao Miyazaki! And with those simple tools, he can create brilliance. Not everyone is talented as Miyazaki. Later in the article, however, he encourages people to become creators, and not simply consumers. With all today’s information overload, it is easy for people to lose sight of what they need to focus on to advance society.

I thought the following quote, included in that Japanese article, was far more illuminating than the “masturbatory iPad use” one. The interviewer asks whether technology like the iPad could be useful in research, offering a way to view information or even to order printed material for delivery. Miyazaki replies (my translation, which suffers from my not knowing what they were talking about up to this point):

Now look. This is going to come across as harsh, but there are some things you just can’t look up on this gadget. It’s because you totally lack an interest in the atmosphere aboard an atake warship, say, or compassion for the men who toiled sweating at its oars. You aren’t going out into the real world and pouring your creativity into something; you’re just skimming its surface, gripping your iWhatever tightly in your hand and stroking away.

I’m sure there are plenty of people who want to put their hands on the latest iWhatever as soon as possible and gain that feeling of omnipotence. Let me tell you something: In the 1960s there were people who went mad for these things called boom boxes—they were huge!—and carried them proudly wherever they went. These people must all be pensioners today, but they were exactly like you. They went crazy for a new product and felt smugly satisfied once they got it. They were nothing more than consumers.

You mustn’t be a consumer. Be a creator instead.

I’m glad Kotaku did include some of this in its English-language presentation of Miyazaki’s views. It’s disappointing, although hardly surprising, that many of the reactions I’m seeing to it on the English-language web is along the lines of “what a goddamn Luddite” instead.

The publication that carried this interview is a Studio Ghibli in-house monthly mag called 『熱風』 (Neppū). The July 2010 issue with Miyazaki’s comments actually dedicated its special feature pages to five articles on the iPad, which makes it hard to view the studio as a rabid den of technophobes.

Looks like an interesting read, and at just ¥2,000 for 12 issues, delivered, I think I’ll be subscribing soon. But here’s where I suddenly shift course and agree that Ghibli is entirely too technophobic for my tastes: to subscribe, you need one of the postal remittance slips included with each issue of the magazine, or, if you don’t have one, you need to send an SASE to the Ghibli folks so they can send you one of the slips. Meh. I’ll stop by one of the bookstores listed on that page to get a copy myself, or maybe just swing by the studio on my way home to cram my sweaty thousand-yen notes into their hands directly.

(Bonus postscript: It looks like there was a similar flap when Gundam creator Tomino Yoshiyuki addressed a video game developers’ conference and told them to stop creating wasteful, “evil” things. There’s coverage of that in English and Japanese, and, of course, plenty of responses from people who ignore the “strive instead to do something meaningful” sentiment in favor of pointing out various hypocrisies in his stance.)

The way around this, according to some threads I found on Apple’s support forums, was to go through the entire iPhone photo library and delete all the offending images, leaving only photos and videos taken with the phone itself. A nice solution if you discover the problem early on, but when you have hundreds and hundreds of photos in there—and they don’t give any indication of which ones are native and which ones were imported via mail or the web—it isn’t such a nice thing to have to deal with.

I did find a nice app called JuicePhone that let me make a browsable backup of everything on my phone. The best part: it didn’t require me to jailbreak the phone, something I have no particular desire or need to do. The other best part: it was free. A nice piece of software to keep around even in this glorious age of iOS 4, when the photo sync issue has been fixed. (I just finished using Image Capture to grab all 3.whatever GB of videos and photos off of the phone and make additional backups of them.)

• iPad is out and is looking to be hugely popular. Publishers are very interested in this new channel for content, and 1.5 million titles had already been downloaded via iBooks as of May 7.
• On May 11 a trade association including some 600 Tokyo booksellers announced its intent to sell digital magazine editions for iPad. Sales will begin in June. Major publishers launched an association in late March to begin work on digital publishing; the trend is picking up steam in Japan.
• President Hagino Masaaki of Voyager Japan, Inc., which produces digital pubs, stated in an interview that when his firm created apps for Kōdansha manga titles and submitted them to the iTunes App Store, around 30% of them were rejected for featuring inappropriate content. App reviewers cited prurient content and scenes of violence and gore as reasons the content couldn’t be allowed to reach viewers through an iPhone app.

So far so good. (Or so bad, if you’re a publisher who wants to get involved here, or a reader who wants to get that content on your new iShiny.) But wait! Today a tweet from @rolandkelts alerted me to this news, though.

Here’s why that doesn’t matter: There are still plenty of multi-comic manga apps on the iTunes store, and every one of them is a mobile reader for a scanlation site. All of them.

This piece spins the situation as “why oh why won’t the manga publishers pull their heads out and offer something that’s just as aggressively priced and easy to use,” but the rejection of all those Kōdansha titles points to a low probability that this could be done.*

Maybe at some point Apple will let publishers provide all their content to readers who take responsibility for their own choice of manga titles, boobies and blood and all. But until that day comes, Apple would prefer that you go download something like Manga Rock (still available on the App Store as I post this) and pirate the scanlations of those same titles. Weird, huh.

* EDIT: In app form, anyway. Maybe your favorite manga will all make their appearance in the iBooks channel and you won’t have to worry about the prudish guys in the App Store office with their itchy reject-button fingers. In that case, this becomes just another story about Apple wanting to sanitize the apps you purchase, only tangentially connected to the fact that some of those apps being sold (or offered for free) today connect you to unsanitized—and blatantly pirated—stuff.

City Poems is the brainchild/retirement project of Victor Keegan, who was a Guardian journalist for 47 years. Keegan, himself a published poet, long ago recognised the potential relationship between poetry and technology—for many years he ran this newspaper’s text-message poetry competition, and his own second collection of poems was published in Second Life. His latest project allows iPhone users to source poems inspired by locations in central London via satellite location; so you could in effect conduct your own poetic tour of the city, following the streets, buildings, statues, buried remains and taverns that inspired the verse.

This seems like the sort of thing that could be done well in Japan. My office is a short walk from the birthplace of Yosano Akiko, for instance; the application could show you some of her verse when you approach the historical plaque on that corner. The whole nation is crisscrossed with travel routes of classical and medieval poets, and dotted with locations boasting rich histories of literary allusion, that could be presented nicely in an app like this.

The big problem is that Softbank’s network coverage is so crappy that if, say, you ride the shinkansen north through Shirakawa, you’ll probably be nearly to Kōriyama by the time your iPhone shows you:

都をば霞とともに立ちしかど秋風ぞ吹く白河の関

I’m not sure what this is supposed to accomplish, really. There’s still at least one online retailer that can deliver you an Apple gadget: Amazon.co.jp marks itself as an authorized reseller and offers the full hardware lineup. (The articles say that Jōshin still sells iPods for delivery, but a quick search makes me think that store is now restricted to accessories as well.)

Perhaps this was meant to strike a blow of some kind against discounters—while the prices at places like Yodobashi and Yamada Denki tend to be the same as Apple’s, you can at least earn points when shopping there to save the 5% consumption tax at the least—but since the brick-and-mortar outlets still sell the machines this hasn’t been stopped.

Apple offers no comment on the situation as of this posting.

None of this affects me much, since I always order my stuff direct from the mother ship (which is the only way to get a US keyboard installed at no extra cost). It’s odd, though, and may be worth keeping tabs on.

(Hat tip @_jm.)

Companies have combined the power of humans and computers to simultaneously double the speed of translation and nearly halve its cost. Where each translator once converted 2,500 words a day at a cost of some 25¢ per word, they can now offer 5,000 words a day at around 12¢-15¢ a word.

Marvelous. This translator makes the same amount of money per day, according to this math, but turns out twice as much text in the target language. Efficiency up, global understanding up. But there are problems here. A few quick, unorganized thoughts:

Problem 1: We aren’t worrying about the fact that this means only half as much time can be spent on proper rereading by the translator and editing by a fresh pair of eyes. The hybrid approach of MT to begin and a human to polish the turds that are MT output means there’s an unhappy person in the mix now—at least I don’t think many people are happy about wrestling with clumsily translated text. I can’t stand it when I’m dealing with stuff a human put together, and even that clumsy human translator is leagues ahead of a machine, and will remain there for the foreseeable future.

Problem 2: The editors who deal with machine output are, ideally, bilingual and capable of doing the translation themselves. If something looks truly odd in your target text, going back to the source text to figure out what’s going on is the only way to set things straight. (Well, there’s actually another way: the monolingual editor just makes a wild guess. I didn’t say it was a good way.) In other words, the ideal form of this man-machine mind meld involves taking a translator who used to be crafting his own sentences and making him clean up the ones a computer spits out at him. Job satisfaction in this new world? Heh.

Problem 3: Don DePalma, chief research officer at a translation outfit, notes that companies need to get their information out there in front of customers in their own languages. “When you’re dealing with anything really expensive or that potentially involves a long-term financial decision—like life insurance or stocks—customers prefer to have information in their own language,” he says. But this is precisely the sort of text that needs to be handled by a specialist, and the companies that sell “really expensive” products will be the very last holdouts still using human pros for the entire process. (It would be fun to see someone trying to market life insurance via Google Translate and an editor in Bangalore, though.) It’s fine to trot this out as proof that companies will need to pay more attention to localizing their material for various markets, but it’s a poor example to bring into the “MT is the future” article.

Problem 4: This.

With [human-assisted machine translation] systems, text is fed into a computer program that tackles the first round of word and sentence conversion using statistics, language rules, or matching with past translations. That covers about 90% of the work. A human then steps in to correct mistakes, clarify sentences, and refine the language for the intended audience or market.

Anyone who’s done translation (at least at a level going beyond churning out crap drafts for rock-bottom prices) or editing knows that the 90% figure here is sheer idiocy. Experienced translators don’t tend to work in phases like this (pump out rubbish at blinding speed and then go back to correct spelling and grammar errors and think about tone and style); they have all these tasks in mind as they go through their text, and it’s hard as a result to define percentages for the effort going into each one of them. But I think the thing that makes translating between human languages a steep challenge for computers is the need to “refine the language for the intended audience or market.” Computers can’t recognize context like that. Humans can, and for human translators, keeping that context in mind and crafting a target text that meets the needs of style, readership, and client preference accounts for vastly more than 10% of their effort. I’d suggest flipping this formula around and saying that the computers handle a tenth of the work, not nine times that amount.

Problem 5: “Language translation is far from being mastered by humans, computers, or any mix of the two.” This is just annoying. It reeks of creationists’ “teach the controversy” demands for equal time for unequal worldviews. Using languages to communicate is what humans do. Birds fly. Fish swim. We talk. What mastery there is in the field of translation belongs entirely to people, and articles like this one need to be written from the perspective of how close computers are to reaching that standard.

Anyway. Enough problems. I’m of two minds when it comes to predicting the future of machine translation. On the one hand, I think the human capacity for language is too deep and too broad for machines to ever take it over completely, and even if 90% of clients end up happy with dirt-cheap mediocrity, the 10% of clients still paying for human quality will represent a healthy chunk of a growing language-services pie. So the good translators will still be making money, and it won’t be by massaging the output of a Google data center.

On the other hand, though, if the scientists ever crack this mystery wide open (perhaps by giving up on computers with nothing but 0s and 1s to deal with and creating new machines that function more like a brain) then we’ll get our translating machine. I’ll be out of a job, along with all my translator and interpreter buddies. But of course we’ll have plenty of company in the unemployment lines, since computers with real thinking power will already have taken over more menial tasks like piloting airplanes, writing software, drafting legislation, teaching children . . .

Handset price: ¥23,040 for 8GB, ¥34,560 for 16GB (paid in ¥960 or ¥1,440 monthly payments over the course of the two-year plan).

Service price: ¥7,280 a month (including the ¥980 White Plan, which includes free calls from 1 a.m. to 9 p.m. to other SoftBank numbers; the fixed-price [unmetered?] data plan for ¥5,985; and the S! Basic Pack, which costs ¥315 and isn’t really described on that page).

Not a horrible deal, all in all, considering what was being predicted for this thing. Still, if you do a lot of telephoning the charges will stack up quickly: SoftBank gives you a great deal on calls to other SoftBank users, but makes you fork over north of ¥20 per minute to all other mobile and fixed-line numbers. Email is free to and from all addresses (you get an @i.softbank.jp address with the thing, but of course you can use all your webmail as usual) and SMS doesn’t exist in this country.

Now to decide whether I really want to ditch the DoCoMo set and jump into the Apple end of the mobile phone pool . . .