Lewis Carroll sive de translationibus

[AL:]

Besides all other perfect qualities, “Alice” (whether in Wonderland, or Through the Looking-Glass) by L. Carroll is a perfect example for those who are interested in understanding translation (using “formal” methods for translation).

Layers and Context #

I’d like to outline a few “layers” of translation here - and each of them can be illustrated with examples from Carroll:

1. Trivial syntactical/semantical one: a rose is “роза”, a cat is “кошка” (or “кот” or “Katze” or “chat).

“Où est ma chatte?” - is our first example from “Alice”.

“So she began again: “Où est ma chatte?” which was the first sentence in her French lesson-book. The Mouse gave a sudden leap out of the water …”

2. Translation of idiomatic expressions. A good example here will be this treatise on character.

A much more difficult exercise, but still accomplishable.

3. Translation of complete nonsense (!). Here is our example: “What’s the French for fiddle-de-dee?”.

Another (non-Carrollian) example would be translation of “Глокая куздра” “Гло́кая ку́здра ште́ко будлану́ла бо́кра и курдя́чит бокрёнка” can be translated as “The glocky kuzdra shteckly budled the bocker and is kurdyaking the bockerling.” Or, perhaps, as “The gostak distims the doshes”.

We can see many examples in Jabberwocky - see “Sample translations” there. In fact, as Humpty-Dumpty explained to Alice - it may not be a complete nonsense at all …

“That’s enough to begin with”, Humpty Dumpty interrupted: “there are plenty of hard words there. ‘Brillig’ means four o’clock in the afternoon–the time when you begin broiling things for dinner.” “That’ll do very well”, said Alice: “and ‘slithy’?” “Well, ‘slithy’ means ’lithe and slimy’. ‘Lithe’ is the same as ‘active’. You see it’s like a portmanteau–there are two meanings packed up into one word.”

And, yet another twist: “Some of the words that Carroll created, such as ‘chortled’ and ‘galumphing’, have entered the English language and are listed in the Oxford English Dictionary.”

Mind you: translating nonsense is very different from translating “Colorless green ideas sleep furiously”, which may seem nonsensical, but on a “higher” level (words do have meaning and syntactically it is a correct sentence). Not only such sentences are trivial to translate, they are almost trivial to interpret (get a meaning) - e.g. as in “The ideas of ecological movement, though bland and inefficent right now, still bother many.”

The moral is: nonsense is translatable if you take it with its context, and, perhaps, Google’s Deep Learning approach does exactly that.

4. Translation of cultural differences - a total switch of the paradigm still leaving the most important features of the original (and who decides what is important?). Check it out here: “Alice in Africe” and others.

Invariants and Methods #

In terms of a layman (and I am one), the problem of translation lies with preservation of a certain invariant, and depending on the “deepness” of that invariant we get very different translations.

Layer 1a.

For the first layer above we can refer to Noam Chomsky and his Universal and Transformational-generative grammars. The idea, essentially, to preserve the syntax using its counterpart in the destination language, and translate the words using words similar semantics.

Layer 1b.

Word translation is not that simple even for the first layer. I believe that the works of Vasily Nalimov can shed some lights on the issue. Regarding meanings of specific words as correlated probabilistic distributions we find the meanings that “fit the best”. This is, essentially, an optimization problem.

Nalimov has an example (in “Вероятностная модель языка.” — М.: Наука, 1974) explaining how [certain] jokes work. The narrator starts on a joke, and goes on, and the listener attributes some meaning to what had been said so far, using most probable/optimal combination from word distributions. But, some of the distributions are bimodal (or worse), and at some point of the narration some other combination suddenly/catastrophically (in the sense of Catastrophe theory) replaces the original one, a we get a comical effect.

As an example (not super-funny) I can quote the joke in which a drill sergeant asks his underlings “Who likes moving pictures?”. You can guess what follows. Here “moving pictures”’s meaning distribution is, at least, bimodal.

A quote:

”…Вероятностный характер языка находит широкое выражение в культуре. Так он проявляется в феномене шуток. Шутка построена на актуализации неожиданного смысла, того, который обычно находится в хвосте вероятностного распределения смыслов. Смешным и забавным представляется нам и уравнивание по вероятности смыслов, которые, как правило, имеют различные вероятностные «веса». Это уравнивание неравного придает пикантности, на вероятностной игре смыслов построены многие французские анекдоты, где одно и то же слово может означать множество самых разных вещей."

Елена Всеволодовна Золотухина-Аболина. В. В. Налимов. 2005.

See also Василий Васильевич Налимов. Вероятностная модель языка. 1979.

Layer 2.

Translation of idiomatic expression usually tries to preserve some principle (or invariant).

In this treatise on character, for example, translators either preserve the original idea

“vinegar that makes them sour” => “уксус делает их кислыми” (a quality of a food item/spice becomes a similar quality of character) - which is pretty shallow, or

uses a new stemmatic principle, preserving a deeper invariant (some food/spice => some character trait):

“vinegar that makes them sour” => “от уксуса — куксятся” (same or similar word stem/root).

A deeper invariant may lead to a better translation, since some of limitations are being lifted.

Layer 3.

Complete nonsence has non meaning per se (by definition), but it does has meaning in its context. Hence, we need first translate the context, and then see what would fit in there. Very few people translated “fiddle-de-dee” as “fiddle-de-dee” - it is too English. And “фу ты, ну ты” seems to fit very well in Russian context. Translation of the context, in this case, seems to be an easy task (layer 1 level).

I believe (I may be wrong though) that BERT works on similar principles.

A quote:

“This characteristic allows the model to learn the context of a word based on all of its surroundings (left and right of the word).”

Layer 4.

Translation of cultural differences is the most difficult task, apparently. Let’s leave alone African Alice - a much simpler example is translation of Carroll’s parodies of well-known verses of his time, such as “Father William” by Robert Southey.

Many solutions were suggested:

• for example, using well-known verses in the destination language (obviously, having nothing to do with Father William). The invariant here is ‘parody for a [some] well-known poem’:

Вечер был, сверкали звезды,
На дворе мороз трещал.
Папа маленького сына
Терпеливо просвещал.

  Пересказ Б. Заходера

• or, translating the parody

Папа Вильям, — сказал любопытный малыш, —
Голова твоя белого цвета.
Между тем ты всегда вверх ногами стоишь.
Как ты думаешь, правильно это?

   С. Я. Маршак

• but, also translating the original poem specially for the occasion (the invariant is ‘parody for the “same” poem; it is not well-known in the destination language. A bummer. So, OK, let us introduce it.’):

Папа Вильям, - сказал любознательный сын, - 
Голова твоя вся поседела. 
Но здоров ты и крепок, дожив до седин. 
Как ты думаешь, в чем же тут дело? 

  Д. Орловская

from

"You are old, father William", the young man cried, 
"The jew locks that are left you are grey; 
You are hale, father William, a hearty old man, 
Now tell me the reason, I pray". 
 
  Robert Southey

and

"You are old, father William", the young man said, 
"And your hair has become very white; 
And yet yon incessantly stand on your head - 
Do you think, at your age, it is right?" 

   L. Carroll

correspondingly.

More examples can be found, for example, here: H.M.Демурова. О переводе сказок Кэрролла.