March 20, 2002

Translation Software Problems

Dear Aunt Eva,
I didn’t imagine myself writing to you, but some of the work that has come my way lately is causing me to wonder if I am really suited to the rewriting business. Let me give you a direct quote, from the CV of a man who is outlining a book he authored:

“This book is giving explanation about the industrial scale in connection with not the explanatory for every so-called article but the history of copyright, and copyright etc. And it enables to understand business laws management, such as copyright, also by the beginner about the overall structure and the law relation of the institutional person concerned in connection with plot and this law.”

This man describes one of his past responsibilities as: ” About the jurisdiction business of the Ministry of International Trade and Industry, it was engaged in research/investigation and the business to study according to the request from a Councilor.”

Aunt Eva, am I going soft in the head or is something happening out there that I haven’t caught up on yet??

Aunt Eva answers:

Dear Evan,

Tally-ho! Bells are going off all over the place on this one. Is it Aunt Eva’s imagination, or does this “translation” have a whiff of those online software translation programs designed to give English speakers access to Japanese web sites. These were sent to Aunt Eva by alert friends who are also sumo fans. Exhibit A is Excite, Exhibit B is Babelfish.

A) * The leaving the hospital ozeki * Chishiro ocean left the hospital. The left foot being able to hurt, it closed the Chishiro ocean from 17 days of Akiba place 9th day, was hospitalized in the hospital inside the Tokyo. There is no hindrance in daily life to to recover, in the future it keeps rehabilitating.

Translation, done by a human being from the Japanese: Ozeki leaves hospital: Ozeki Chiyotaikai has left hospital. After injuring his left leg he was absent from Day 9 (September 17th) of the Aki Basho (Autumn sumo tournament), but he has recovered sufficiently to have no trouble in daily life. From now on he will be in rehabilitation.

B) Koto Koki of the bill debut called “three Nihon University glass” by the spring sumo tournament for 99 years, ??, and the peak of your opinion were broken continuously, and it was brought into the limelight. However, after that, it is not torn but looks back upon the wall of Makushita, saying, “The time which a program does not go up even if it trains was the most painful.” Change of presence also decreases now and he is disguised as the sumo wrestling which comes out to a front. There is the line training effect declared “It is a hobby”, and activity by juryo is also expected.

Human intervention proved fruitless on this one, as the Japanese original could not be found.

So it seems, Evan, that your prospective client has fed his CV into some translation software and thinks that it only needs a bit of tidying up by a native checker. If you can get the original Japanese, you may be able to retranslate (but are you being paid enough?); even more time-consuming is translating each word or phrase back into Japanese and then putting them together differently; but again, is it worth it? If you possibly can, tell him to scrap this, print out the original Japanese, and give it to a competent translator.

The greatest danger of this translation software is that clients who are not translators will believe the blurb. At certain levels and in certain language groups, it works quite well. But with language pairs like English and Japanese, which share no common foundation, it tends not to work at all for more than a few regular phrases. Even a much simpler technology like transcription is hard to put into practice for Japanese because of the great number of words that sound the same. (Many parliaments around the world are keeping minutes in this way; the technology here is in its infancy and for this very good practical reason.)

Now we’ve opened this particular can of worms, let’s look a bit further and try to identify the new problems that arise when we utilize the new technology that is laid so invitingly at our fingertips. Error analysis is nothing new. For instance, in the pre-printing days when books had to be copied by hand, we can tell that a book was the work of an individual scribe if he accidentally skipped a paragraph that began with the same words as the next; in other words, the errors are visual. In ancient Rome, some of the big booksellers had a mass-production system. How do we know? Common errors in books produced this way are mistakes of hearing, similar-sounding words confused. Obviously one person was dictating the book to a roomful of scriveners.

What are the modern equivalents, and how can we advise people to avoid them?

The simplest is represented by: “I have the pressure of writing to you to seek your support. . . .” This stems from over-reliance on the spell-checker, compounded by an r/l problem; and, of course, the writer doesn’t read her finished work—or else sees it through rose-colored spectacles. Aunt Eva learned early to ignore mistyping that produced impossible words, because the spell check software would catch them later; but if she accidentally typed a real word that happened to be the wrong one (e.g. weekly/weakly), it must be corrected immediately because she might forget it later and the spell-check would let it pass.

Another bad thing that happens to rewriters just because it’s so easy for the person who does it: an identical letter is to be sent to multiple recipients. Requiring a hanko on each, the secretary prints out, say, ten copies, differing only in the addressee, and delivers them all to the native to be individually checked. What is to be done here? Well, if the native is in-house and being paid for a full day’s work, he/she might as well check all ten. Money for jam. If, as is more likely, the native checker is a part-timer or on retainer, the secretary has to be trained to print out one copy of the letter for checking and a list of addressees. Otherwise, Aunt Eva’s jaundiced experience has been that one of two things can happen: if she corrects the top copy only, and writes, “the rest are the same,” it is very likely that one correct letter will be send out, along with nine uncorrected ones; or (heaven forfend!) ten copies of the letter will be sent to one addressee, while the other nine, like the littlest little piggie, get none.

Retrospective Aunt Eva, going through her “Work Problems” file for more ammunition realizes that she herself had failed to identify this problem earlier. That “it” as a subject is the first giveaway that automatic translation is at work, because it is axiomatic in Japanese academic circles that “An English sentence must always have a subject, whereas Japanese often omits it.” But as any native speaker knows, the subject can often be omitted in English too; not least in the format used for CVs: “Graduated from AAA university, entered ZZZ Ministry.” The ellipsis that comes naturally to us is also characterized as “poor English” by those who have learned the language, however thoroughly, in Japan.

So a Japanese who gets this kind of output from translation software will not be aware of a strong smell of fish: “What kind of things are staff member’s concrete duty contents? It is illustrated concretely, and it wants it to teach.” An experienced rewriter can easily turn this into “What is the content of staff members’ duties? (please give as many specific examples as possible)”; but it is surely beyond the job description.

Another problem exacerbated by the computer is the production of new “boilerplate,” that body of already proofed, perfect text that can be copied and pasted with confidence into a new document. Right? Wrong. Take the gilt-edged invitation card bearing the heading, “XXX has the pleasure of the company of. . . .” A line has been dropped from the pattern invitation and repeated for years.

And, in the opposite direction, the superiority of the computer over the typewriter is sometimes negated by someone who is still wedded to her typing skills; if something is dropped in a cut-and-paste boilerplate section, it is quicker to hastily type in the missing part, but the checker may not be aware that a piece already proofread has been tampered with in this way. Aunt Eva recalls with sorrow and shame letting through an unusual part of the divine anatomy that featured in the Missa Brevis: “Agnus Dei (Lamb of God)” had mysteriously changed into “Anus Dei.”

Quickies:

1) What flavors will they think of next?

The parlor serves 20 flavors of ice cream in both paper cups and cones that can be complemented by 20 toppings, such as fruit pieces, carmel sauce and chocolate moose. (Daily Yomiuri advertising blurb, 3/7/01)

2) How not to use a dictionary

If a new book is not available, hackneyed is possible. (Request from a French researcher)

New translator with degree from 4-year US college: I am sorry for make you busy. Those documents have to hand in by Friday.

Now that phrases can be stored, problems of frequency will become even worse: e.g., ‘as well as’, ‘furthermore’, ‘moreover’

Some things will never change: The G of J expreses its deep sorrow over the people died during the operation against this terrorist acts and sympathy for bereaved family.