Lost in Autotranslation

[lpetrich]

The Shallowness of Google Translate - The Atlantic -- about natural-language translation software. Author Douglas Hofstadter objects rather strongly to the proposition that the existing sort of autotranslation software will make human translators pretty much obsolete, except for doing touch-ups and the like.

I don't feel as strongly, but I think that his overall point is correct. Rhythms in words and other such things can be difficult to get across without some creative less-than-literal translation. I also find that autotranslators often do clumsy-looking translation, and sometimes very screwy trasnlation.


Sometimes it is fairly easy. "The Song of the Bell" (Friedrich Schiller, 1798) contains this inscription on the titular bell, an inscription in Latin:
Vivos voco
Mortuos plango
Fulgura frango

A fairly literal translation nevertheless keeps the rhythm of the original:
I call the living
I mourn the dead
I break the lightning


A certain Alexander Unzicker once wrote a book in German, "Vom Urknall zum Durchknall". A literal translation of that title is "From the Big Bang to madness", but that title lacks the rhythm of the original. I thought of a less literal but more rhythmic translation: "From the Big Bang to the Big Crack-Up". He and Sheilla Jones worked on an English translation, but they titled it "Bankrupting Physics".


I close with noting an old Italian saying, "traduttore, traditore": "translator, traitor".

 

[lpetrich]

In Lucius Apuleius's novel "The Golden Ass" or "Metamorphoses", the hero gets turned into a donkey in the literal sense, and he gets used as a beast of burden by several people. Among them are the eunuch priests of a Syrian version of Cybele. They had an oracle that they interpreted for everybody's circumstances (Bk. 9, Chap. 8):

Ideo coniuncti terram proscindunt boves,
Ut in futurum laeta germinent sata.

Word for word:
Therefore/for-that-reason joined-together(pl) earth(obj) plow-up(3p) oxen,
So-that in future abundant grow(3p-subj) crops/standing-grain.

Robert Graves (1950):
Yoke the oxen, plough the land,
High the golden grain will stand.

With accented syllables:
YOKE the OXen, PLOUGH the LAND,
HIGH the GOLden GRAIN will STAND.
(Not very literal, but with a nice rhythm)

Sarah Ruden (2011):
An ox team cleaves the soil for just this reason:
Sown grain will sprout abundantly in season.

Joel C. Relihan (2007):
For this the team of oxen plows the furrowed earth,
So fertile fields of grain will sprout in times to come.

E.J. Kenney (1998, 2004):
The yokèd oxen drive the furrow now,
So that one day luxuriant crops shall grow.

Thomas Taylor (18th cy.?):
Together yok'd the oxen, till the ground,
That stocks abundant may from thence arise.

William Adlington (1566, 1639):
The Oxen tied and yoked together,
Doe till the ground to the intent it may bring forth his increase.

 

[fast]

The translations are decent for normal straight forward messages, but the more odd, esoteric, or crafty the message, the more garbled the output will be. Like he said, there's no understanding.

 

[Copernicus]

The quality of automatic translation varies with the target language, the subject matter, and the sophistication of the software. Google Translate is actually quite good for some languages, especially European ones. For others, it can be absolutely atrocious (e.g. Turkish). However, I have found that it is a very nice supplement to my comprehension with foreign languages that I have some fluency in (e.g. Russian, French, Spanish, Italian, German). I have never studied Portuguese, but I recently found it to be amazingly good, given my familiarity with other Romance languages.

Several years ago, I attended an interesting seminar on machine Translation from Yorick Wilks, a well-known expert in the field. He discussed the success of low-quality literal translations and pointed out something that is very true--that there is a huge market for even bad translations. Machine translation is a lot cheaper than human translation, and sometimes, you just don't need 100% accuracy. Nevertheless, automated translation has improved by at least an order of magnitude since then. Professional translators rely on translation software for obvious reasons--it speeds up the amount of text that can be translated and improves accuracy. Translators who don't use the software simply cannot compete. So machine-aided (vs full automatic) translation is a strong driver in the popularity of translation software.

However, content is paramount. If the language is fairly predictable (e.g. weather reports), then the quality of translation can be of such good quality that it doesn't make sense to have human post-editing. It if is varied and full of metaphors and idioms (e.g. poetry, literature), then human translators are essential.

 

[Hermit]

Automatic translators would perform quite poorly tackling something like any of Shakespeare's plays at least for now. Laid up in bed for what seemed like an eternity I had an idea to banish boredom. I pulled some Shakespeare play and a German translation of it from my parent's bookshelf and did a side by side comparison of some of it. I can't remember now if the translator was Schiller or Schlegel. This was almost 50 years ago.

In any case, the translation was a work of genius. I was continuously surprised and astounded how the problems of idiom, rhyme and rhythm were overcome while staying faithful to the style and spirit of the bard. Mysteriously, I am tempted to say miraculously, the two texts managed to be simultaneously different and the same.

The rest of my bedridden days flew. They turned into one of the best journeys of my life.

In principle some future algorithm will probably produce similar results. Eventually. Machine translation has a long way to go.

 

[Copernicus]

Machine translation works much better now than in the past, because statistical language methodologies have really improved the ability of machines to pinpoint intended word usage, as long as the language is fairly literal. The way they work is that they look at a large "window" of text, and they choose the intended meaning of a single word on the basis of how frequently it appears with that type of word cloud. Such techniques perform poorly on literature and poetry, because those forms of writing are full of metaphors. Newspaper articles, however, tend to rely mostly on literal word usage.

The standard joke in the field of machine translation is that an early program (Systran, I believe) was tasked with translating the sentence "The spirit is willing, but the flesh is weak" into Russian. Then it translated the Russian translation back again into English. The English that came out was “The vodka is acceptable, but the meat has spoiled.”

 

[Wiploc]

Thomas Taylor (18th cy.?):
Together yok'd the oxen, till the ground,
That stocks abundant may from thence arise.

I really want to move that comma:

Together yok'd, the oxen till the ground,
That stocks abundant may from thence arise.

 

[Wiploc]

I wrote a limerick. Just for fun I put it into BabelFish, and translated it into a language and back:


German:

It was a role model, Ulysses.
They were free after his mistress.
He lied and bolted;
He was violent and rude.
A history of morals is what Thisis.


French:

There is a role model, Odysseus.
The contenders have been after his missus.
He lied and screwed;
He was violent and rude.
A moral tale is that Jose.


Spanish:

There was a model to follow, Ulysses.
The suitors were after his missus.
He lied and bolted;
He was violent and rude.
A moral is that you Thisis.


Greek:

There was a role model, Odysseus.
The suitors were after the Kira.
Lied and screwed?
It was violent and rude.
A morality tale is what this is.


Hindi:

It had the role model, Odysseus.
The suits were followed by Kira.
Lied and grated?
It was violent and rude.
A morality story is what it is.


Italian:

There was a role model, Ulysses.
The suitors were after his wife.
He lied and screwed;
He was violent and rude.
A morality tale is what this is.


Chinese:

There is an example, Ulysses.
The suitor is after his wife.
He lied and screwed up.
He was violent and rude.
That's the moral story.


And of course the game of telephone where we go from English thru all of the above languages, and then back to English:

Mechanism, Odysseus, followed the pattern.
They were free after her boyfriend.
He lied, bolted,
and he was violent and rude.
It's a story about morality, not Joseph.


What I conclude as a result of this exercise: Some of those alphabets have funny looking letters.

 

[Loren Pechtel]

I broke Google with one word:

Cowed

It came back with the animal that goes moo.

(I'm sure what it did is saw that "cowed" is the past tense of "cow"--but then picked the noun meaning (the most common one) rather than the verb.)

 

[LordKiran]

https://www.waverlylabs.com/

Seems appropriate.

 

[lpetrich]

Titled link: Waverly Labs: No More Language Barriers -- it's autotranslator headphones and an autotranslator app for smartphones and tablets. So autotranslators can now do spoken language.

I also used Google Translate on

Vivos voco.
Mortuos plango.
Fulgura frango.

Ideo coniuncti terram proscindunt boves,
Ut in futurum laeta germinent sata.

and it gave me

The quick call.
Mourn the dead.
The lightning break.

Therefore, the composite ground raile cows
In order to be happy to germinate crops.


The first one get the verb conjugations mixed up. Translating back into Latin gives:
Vivi vocant.
Mortuos plangite.
Fulgur frangit.

The second one is even worse.

 

[lpetrich]

There's a Eurodisco song whose main words are

Eins, zwei, Polizei
Drei, vier, Grenadier
Fünf, sechs, alte Keks
Sieben, acht, guten Nacht

Literal translation:

One, two, police
Three, four, soldier
Five, six, old biscuit
Seven, eight, good night

In the original, the last word or phrase had 3 syllables.

Nena's song 99 Luftballons was translated as 99 Red Balloons, so as to make 3 syllables and not 2.

 

[lpetrich]

The issue of music brought this to mind for me. In his book Gödel, Escher, Bach: an Eternal Golden Braid, Douglas Hofstadter considers various things that artificial intelligence might possibly do.

Question: Will a computer program ever write beautiful music?

Speculation: Yes, but not soon. Music is a language of emotions, and until programs have emotions as complex as ours, there is no way a program will write anything beautiful. There can be "forgeries” shallow imitations of the syntax of earlier music - but despite what one might think at first, there is much more to musical expression than can be captured in syntactical rules. There will be no new kinds of beauty turned up for a long time by computer music-composing programs. Let me carry this thought a little further. To think - and I have heard this suggested - that we might soon be able to command a preprogrammed mass-produced mail-order twenty-dollar desk-model "music box" to bring forth from its sterile circuitry pieces which Chopin or Bach might have written had they lived longer is a grotesque and shameful misestimation of the depth of the human spirit. A "program" which could produce music as they did would have to wander around the world on its own, fighting its way through the maze of life and feeling every moment of it. It would have to understand the joy and loneliness of a chilly night wind, the longing for a cherished hand, the inaccessibility of a distant town, the heartbreak and regeneration after a human death. It would have to have known resignation and worldweariness, grief and despair, determination and victory, piety and awe. In it would have had to commingle such opposites as hope and fear, anguish and jubilation, serenity and suspense. Part and parcel of it would have to be a sense of grace, humor, rhythm, a sense of the unexpected-and of course an exquisite awareness of the magic of fresh creation. Therein, and therein only, lie the sources of meaning in music.

If one wants to go that route, it would not be enough to experience those emotions. The AI system ought to have our brain wiring, one that makes music provoke emotions.

But failing that, one can fake it by imitating existing styles or else experiment with human subjects to see what emotions the system's compositions provoke.

That is likely true of other art forms, including language-related ones like creative writing and poetry -- and what may be called creative translation.

 

[bilby]

There's a Eurodisco song whose main words are

Eins, zwei, Polizei
Drei, vier, Grenadier
Fünf, sechs, alte Keks
Sieben, acht, guten Nacht

Literal translation:

One, two, police
Three, four, soldier
Five, six, old biscuit
Seven, eight, good night

In the original, the last word or phrase had 3 syllables.

Nena's song 99 Luftballons was translated as 99 Red Balloons, so as to make 3 syllables and not 2.

That latter song's English version is an excellent example of a translation that has to choose between scansion, tone, and literal accuracy. The translator (correctly IMO) sacrificed the third in favour of the other two - for example, keeping the Jet Pilot/Captain Kirk reference in both versions required it to move to a different stanza; and the adoption of the 'Red' motif that was required for scansion to also reference the Soviet bloc in the English version (a reference not made by the German version).

Both songs tell the story of a third world war; The German version is explicit in that this was caused in large part by a desire for personal advancement on the part of the politicians who expected to profit from winning; While in the English version, the events are attributed more to the consequence of a rigid command structure that couldn't be prevented from overreacting to a perceived threat. I get the strong impression that this was an accurate response to the zeitgeist in the two countries at the time - West Germans felt disenfranchised by politicians who cared little for their people, and would sell them out for a profit, while the British and Americans were more concerned with the military machine being increasingly automated and removed from political oversight. The sense of time is also very different; The German version explicitly speaks of 99 years of war, while in the English version the sudden drop in tempo and dynamic between the last two verses seems to imply a short but intense (nuclear) exchange, in the absence of an explicit time-frame given by the lyric.

To fit these two slightly different stories into the same tune, with a similar feeling but sensitively adjusted for the culture of the target audience, while keeping the same overall tone, is IMO genius. It will be a LONG time before a machine could achieve that quality of result when called on to translate from one language to the next - the literal translation of the lyric in either direction would not produce a singable result, much less a song good enough to be a hit in both languages.

 

[lpetrich]

The Dao De Jing (Tao Te Ching) presents a translation challenge. It is written in very terse Classical Chinese, and the translations I've seen seem much more verbose than the original. Let us look at the first few lines of it. The original Chinese characters, a pinyin transcription, a word-for-word translation, and my attempt at composing sensible English.

道可道，非常道。名可名，非常名。

dào kě dào, fēi cháng dào.
míng kě míng, fēi cháng míng.

way be-able way, not constant way.
name be-able name, not constant name.

The Tao that can be traveled on is not the eternal Tao.
The name that can be named is not the eternal name.

 

[Rhea]

I have several friends and relatives who post in not-english and most times translate works, but sometimes it breaks down completely. Like when one German cousin told me to give my son “a thick wet kiss”. Funny, but pretty wrong. Probably should have been more like “Give that kid a big slobbery kiss for me!”

But today’s was a real stumper. From a cousin typing in Hebrew and I can’t make heads or tails of it. This is a dialogue between a parent and a child. (“Open” is what translate thinks the child’s name is)

Open sentences:
We :״ master open you....״
Open: " mode!״
We: " and....״
Open: ״.... ţyk!״
We: " right, and....״
Open :״.... Nectar!״

 

[lpetrich]

I've seen some creative translation in translations of Aristophanes's The Clouds.

At one point, Socrates notes that there are separate words for each sex of domestic animal, though there isn't one for chicken. So he proposes some. I'll give some translations that I once saw:

chicken - alektruôn
male (rooster) - alektôr - chicker
female (hen) - alektruaina - chickeness

Socrates also notes that a man named Amynias turns into Amynia when one calls for him, thus changing his sex. That's the vocative-case form. I once saw a translation that did Alexander - Sandie, using what may be described as an informal English vocative.

 

[lpetrich]

I decided to try translating Robert Graves's translation back into Latin.

Yoke the oxen, plough the land,
High the golden grain will stand.

Bovês conjuge, terram arâ,
Altum stâbit frûmentum aureum.

The oxen we must yoke/join, the land we must plow,
High will stand the golden grain.

The original, with long vowels annotated with hats:

Ideô coniunctî terram proscindunt bovês,
Ut in futûrum laeta germinent sata.

 

[steve_bank]

Well, hairsplitting and grammar checking in editors eliminated dictionaries and grammar books?

Te used to be a professional category of workers who reviews documents for spelling and grammar.

Before Windows Office Suite in companies there were skilled categories of workers who created documents and presentation. Photographers and graphic artists.

It could take weeks ad a group of people to create overhead slides for formal presentation.

Today take a picture with a camera and import it into a document.

The position of typist is long gone. You used to give a hand written document to a typist followed by a review and correction process that could take days to weeks.

Today most do their own typing.