December 09, 2007

The Etiology and Elaboration of a Flagrant Mistranslation

[Guest post by Victor Mair]

A series of earlier Language Log posts have discussed the curious phenomenon seen in the grocery-store sign on the right: absurdly crude English mistranslations in bizarrely inappropriate contexts.

In "Gan: whodunnit, and how, and why?" (5/31/2006), I explained one of the sources of this phenomenon: several Chinese characters pronounced GAN1 or GAN4 -- and meaning such widely disparate things as "dry," "calendrical sign," "to do," and much else beside -- all got collapsed into one simplified character: 干. This has led to enormous confusion, especially when people who know next to no English rely on machine translation software to convert Chinese into English. The chaos caused by this combination of circumstances is vastly exacerbated by the fact that this little, three-stroke symbol also has a vulgar meaning when pronounced in the fourth tone, GAN4, namely "fuck," which is probably an extension of the regular sense of "do." Because GAN4 ("do") and GAN1 ("dry") are now both written with that little, three-stroke character, the damage is compounded by the enormous range of intended senses of GAN1/4 ("dry," "do," "act," "work," "undertake," "shield," "have to do with; be concerned with," "edge of a body of water," "be rude, impolite, blunt," "embarrass or annoy," "give the cold-shoulder to," "empty, hollow," measure word for a group of people, "trunk, stem, main part," "cadre," "competent, capable, able, talented," "go bad," "be a disaster," etc.), all of which are capable of coming out of the translation software as "fuck."

People who see signs employing the f-word all over China, even in large stores and fancy restaurants, are not only aghast, they wonder how the dickens such a gross mistranslation could have originated and proliferated. I believe that the explanation given in the previous paragraph adequately and accurately accounts for the origin of the basic GAN1/4 = "fuck" mistranslation. The question that remains, then, is how did this virus spread? Theories abound, to say the least. I have intelligent colleagues who believe that naughty people do this on purpose just to scandalize customers and clients. Others hold that it is done to make English look like an uncouth language. Still others maintain that devious foreign translators plant these mistranslations all over the place to make Chinese look stupid and crude. And there are so many additional theories that attempt to account for the GAN1/4 = "fuck" misrendering that I can't keep track of them.

Most of us, however, have all along suspected that this phenomenon resulted from reliance on faulty translation software. Indeed, it is easy to prove that absurd English translations are being spewed out daily in China when individuals who don't know English merely plug Chinese sentences into the software and expect it to come up with reasonable renditions.

For example, one of my students has told me about a sign in Tai'an, in Shandong province, near Qufu, the legendary birthplace of Confucius. The Chinese text reads "Liang2shi4 zheng3gu3 tui1na2", which means "Liang Family Bone-Setting (Medical) Massage". (For a description of TUI1NA2 in traditional Chinese medicine, see the Wikipedia entry.) The English text on the sign reads "Whole Bone of the Beam Surname Pushes to Take"!

After trying for more than a year to find proof that the GAN1/4 = "fuck" mistranslation was indeed the result of relying on poor translation software, I am now able to demonstrate that this really does seem to be the case. Here is the evidence.

There is a bulletin board called Lequ Yuan (which means something like "Pleasure Garden"), and on that site there is a page called "Lüyou Wenhua Tan" (which punningly means both "Donkey Friends Forum" and "Travel Friends Forum"). On January 14, 2003 at 00:51:23, somebody calling him/herself "kailash" made a post entitled "Gaoxiao fanyi (Ridiculous Translations)." The contents of kailash's post (in the top left) are the following, in my English translation:

Truly ridiculous!

Location of photograph(s): The new X supermarket at Hexi in Changsha (Hunan).

The one that is most classic is that for GAN1HUO4 [= "dry goods" -- it must have actually been "fuck foods" in the photograph, which is no longer accessible on the pku.edu.cn web site]. Can GAN1 really be translated that way?

Below are some (translated) comments on that post, with the commenter's name after each one:

"I suspect that somebody's been playing tricks." Balang

"You go buy a Jinshan Ciba with Added Jinshan Kuaiyi." Huayuan ("Flower Garden")

"Really!" kailash

"I still don't believe it." windy

"Ha, ha, ha!" Knight yvan

"Is it true or fake? Somehow I suspect that it has been faked." By No Means a Lonely Ostrich

Huayuan followed up with another post later that same day, at 11:05:21 to be precise. His response to the doubters, begins thus: NI QU MAI GE JINSHAN CIBA JIA JINSHAN KUAIYI ("You go buy a Jinshan Ciba with Added Jinshan Kuaiyi").

Now, just what is this Jinshan Ciba that can be equipped with a Jinshan Kuaiyi? Literally rendered, these names mean "Gold Mountain Word Hegemon" and "Gold Mountain Fast Translator." Jinshan Ciba is an electronic reading aid (a sort of glorified bilingual [Chinese and English], and to a certain extent multilingual [since some Japanese is included], dictionary), while Jinshan Kuaiyi is a machine translation device. These tools are widely used all over China. Nearly all of the PRC students at the University of Pennsylvania have brought theirs with them, although they realize that Jinshan Kuaiyi usually produces English that is laughable.

Huayuan's post continues (emphasis added):

Then let it (Jinshan Ciba / Kuaiyi) translate this sentence: NI3 XIANG3 GAN4 SHEN2ME? ("What do you want to do?")

It will come out with: "what do you want to fuck?"

This is an experiment that can be repeated; scientifically it is sufficiently rigorous.

Now, that was in 2003. I have checked the latest version of Jinshan Ciba / Kuaiyi, and it does not return the vulgar mistranslation recorded by Huayuan, although it still comes out with a lot of other gibberish (see below). It would appear that somebody told the makers of Jinshan Ciba / Kuaiyi about this gross problem with their translation software and they have taken steps to remedy it. However, the damage has already been done and the cancer continues to spread throughout the Chinese body linguistic. There still must be countless earlier versions of the Jinshan software out there that are being used daily. The only way to halt this ludicrous phenomenon is never to paint, print, or publish an English translation without first checking with someone who is fluent in the target language. That, of course, is unlikely in the foreseeable future, because there just aren't enough skilled speakers of English to go around for the huge demand that has engulfed China.

I tested the latest version of the Jinshan software, and here are some samples of what it now produces (screen shots taken a few days ago):

NI3 XIANG3 GAN4 SHEN2ME?
Jinshan: "What do you want to do?"

NI3 XIANG3YAO4 GAN4 SHEN2ME?
Jinshan: "What do you want to do?"

NI3MEN ZHENG4ZAI4 GAN4 SHEN2ME?
Jinshan: "You just at stem what?"
corrected translation: "What are you doing right now?"

NI3MEN GAN4 SHEN2ME NI?
Jinshan: "What do you do?"
corrected translation: "What are you doing?"
or "What do you think you're doing?"

NI3MEN GANG1CAI2 GAN4 SHEN2ME?
Jinshan: "What are you just dry?"
corrected translation: "What were you doing just now?"

NI3 GAN4 SHEN2ME ZHE4ME SHUO1?
Jinshan: "What do you do to say so?"
corrected translation: "What do you mean by talking like this?"
or "Why are you speaking this way?"

GAN1HUO4
Jinshan: "Dry goods"

GAN1TIAO2 QU1
Jinshan: "The stem adjusts area"
corrected translation: "Dry Seasonings Section."
(see below for a brief discussion)

Those who travel around China know for certain that the GAN1/4 = "fuck" abomination is not fake. It is real, and now -- though it remains thoroughly deplorable -- I think I understand how it happened.

Here are just two of the countless instances of the GAN1/4 = "fuck" paradigm that have spread throughout in China:

Pinyin:     GAN1TIAO2 QU1
English: "Fuck to adjust the area"

Correct translation: "Dry Seasonings Section"

Note: It is no wonder that machines get confused by this expression (see the less salacious machine translation above, no. 8 of the screen shots, "the stem adjusts area"), since every Chinese to whom I've shown this sign has hesitated in their pronunciation (the second syllable could also be read DIAO4) and in their interpretation of its meaning -- there are many different possibilities.

Pinyin:    SAN3 GAN1GUO3
English  "spread to fuck the fruit"

Correct translation: "Loose Dried Fruit"

Note: More often GAN1GUO3 actually refers to nuts instead of fruit.

Many other examples of GAN mistranslated as "fuck" can be found on the internet -- several have been featured in the earlier Language Log posts linked above, for example this sign from a hotel buffet:

And this restaurant sign:

Two more: what should be "dry foods price counter" is rendered in large letters as "FUCK THE CERTAIN PRICE OF GOODS":

And what should be "dried foods" becomes "FUCK GOODS":

As a sample of the other widespread effects of unwise reliance on dictionaries, digital or otherwise, here are some incredibly fine examples of Chinglish in Shanghai., probably also caused by excessive reliance on poor-quality machine translation. Among other precious items, notice the "Bang products" and the "antithetical couplet ocean."

Well, enough for now. I hope that I do not have to spend the rest of my life documenting and explaining Chinglish. For the moment, however, it would appear that I still have much work to do.


I wish to express my gratitude to Jonathan Smith for helping me track down early occurrences of the GAN1/4="f*ck" atrocity on the Web and for the fabulous TUINA sign in Shandong, and to Jiajia Wang for making the screen shots of Jinshan Ciba / Kuaiyi in action, also for scouring Facebook for interesting examples of Chinglish.

[Above is a guest post by Victor Mair]

[Update -- Joel Martinsen writes:

I dug up a CD of Kingsoft 2002 and tested it on some of Prof. Mair's example sentences. Thought it might be nice to have a screenshot of how the software used to work to compare to the more acceptable translations of recent versions of Kingsoft QuickTrans:

The last line of the attached image is part of the translation of the title of an academic paper that was published in the Economic Administrative Cadre Bulletin in 2003, and discussed here in the EastSouthWestNorth blog:

* [019] A Solid Fuck (12/08/2007) (KDnet) In issue 2, volume 16, June 2003 of the Kansu province Economic Administrative Cadre Academy Bulletin, there appeared an essay titled <开拓进取,真抓实干,不断开创西部大开发的新局面>. The author Song Chaosu is the Kansu provincial party secretary and the chairman of the Kansu provincial people's congress standing committee.

In the database, the title of the article was translated into English as: Expand Enterprising and Really Grasp Solid Fuck and Continuously Expand and Great the New Situation of Buildings of Western Region.

If there is a "solid fuck," it is the translation job. A more appropriate translation is: Develop and Forge Ahead with True Understanding and Effort in order to Continuously Create New Situations to Open Up the Western Region."

]

Posted by Mark Liberman at December 9, 2007 06:59 AM