Week 7: Translation/Localization
Checklist
Bellos on translation (3 chaps.)
Aykin and Milewski, "Practical Issues"
At-home Exercise: Machine translation exercise
In-class exercises and at-home assignments
At-home Exercise: Machine Translation
Since my Hindi isn’t too strong anymore (10 years since I last studied it) — I decided to translate some famous couplets by Kabir Das, a 15th-century Indian mystic poet and saint. These couplets are often quoted in situations where a proverb might be used — giving advice or some wisdom about life. These are often repeat by parents, much to the annoyance of their children (I can attest, from personal experience :P)
सुख मे सुमिरन ना किया, दु:ख में करते याद ।
कह कबीर ता दास की, कौन सुने फरियाद ॥
”Did not sum up in happiness, remembering doing in sorrow.Say Kabir Ta Das, who has heard the pleaad.”
रात गंवाई सोय के, दिवस गंवाया खाय ।
हीरा जन्म अमोल था, कोड़ी बदले जाय ॥
”The night is lost, the day is eaten.Diamond birth was Amol, Cody would be changed.”
दुःख में सुमिरन सब करे सुख में करै न कोय।
जो सुख में सुमिरन करे दुःख काहे को होय ॥
”Do everything in happiness without doing anything in happiness.The one who comforts himself in happiness, is in pain.”
बडा हुआ तो क्या हुआ जैसे पेड़ खजूर।
पंथी को छाया नही फल लागे अति दूर ॥
”What happened if it grew like tree dates?The sect was not shadowed too far”
Analysis
Did MT translate any of your text correctly? If so, what?
It got certain words right, but barely any of the grammar. There were also several words that were completely wrongly translated, or not translated at all. Hindi is a very high-context language, since one word can have several varied meanings that are often not even related to each other.
Assuming that your poem contained some poetic language, allegory, alliteration, symbolism and/or emotional ideas, what concepts in the source were completely missed by the MT output? Show some examples of this in your report.
3rd couplet — the translation almost inverts the meaning of the entire couplet. The translation would be: “Everyone prays during bad times, no one prays during good times. But if you prayed during good times, there wouldn’t be bad times.” However, from the machine’s translation, the last line “the one who comforts himself in happiness, is in pain” almost implies that you should pray in happiness because then you will sad? Weirdly the exact opposite of what the couplet is saying.
4th couplet — the English translation makes no grammatical sense at all. What the couplet is really trying to say is: “So what if you’re big/tall like a date tree? You’re of no use to a traveller because you provide shade.” The deeper meaning being: “Size or might isn’t everything; utility and function are more important.”
How do you think the translation could be improved?
I’m not really sure how Google Translate works exactly, but I imagine it uses a dictionary to look up words and do a word-by-word translation. It then probably applies some grammar rules at the end, but the grammar rules (especially those concerning order of words are quite fuzzy in Hindi). So I imagine to improve the translation, it would help if there were some existing translations it could “learn” from or cross-reference at least. I.e., machine learning or artificial intelligence. Since these were poems, it’s extra hard to translate — even for a human.
It’s really difficult to ensure that metaphors are not lost when we translate things from one culture to another—poetry is really difficult to translate for human beings. Any chance an MT engine could ever get this right, based on the context and content of this poem?
I think it would be quite impossible for MT to translate couplets like these accurately. These couplets say one thing on the surface, but are referring to something completely different on a deeper level.
Quotes
Aykin: Practical Issues
This is a good chapter to have as a reference when doing some of the translation or internationalization work required of designers.
“The quality of translation is highly correlated with how well and clearly the original text is written.”
This was an especially good point, that applies to the entire design process as well. If your original design is conceived with internationalization in mind, then it will be easier to do so, and the results will be better too.
Bellos: Is That A Fish In Your Ear?
In general I liked these chapters, but had a tough time linking them directly to UX. I would definitely read the rest of this book though, seems interesting.
“The evidence itself brought him to see that any attempt to match the grammar of a language with the culture of its speakers or their ethnic origins was completely impossible. “Language,” “culture,” and “race” were independent variables.”
Bellos doesn’t go into too much detail about how Sapir came to this conclusion, but I think Sapir meant it like: you could take an individual from any “race” and teach them another language or immerse them in another culture and they could be perfectly competent at it — hence implying that those are independent variables. However, I’m not fully convinced of this, since growing up with a certain of those three variables would definitely impact the other two. They may be independent of each other, but there is definitely a confounding/hidden variable that links them.
Anyway, I guess I’m a little confused about this part, since a lot of the other readings we’ve been doing claim that language and culture are inextricably linked.
“What distinguishes translating UP from translating DOWN is this: translations toward the more general and more prestigious tongue are characteristically highly adaptive, erasing most of the traces of the text’s foreign origin; whereas translations DOWN tend to leave a visible residue of the source, because in those circumstances foreignness itself carries prestige.”
I could see this concept being applied when translating or internationalizing websites and apps, but I wonder whether it would be “ethical” of designers to assume that one language is more “prestigious” than another and let that assumption influence their translation decisions. I can imagine two very different outcomes if designers assume they are translating UP/DOWN vs. translating laterally. Should the audience get to choose what translation they receive, or should the translators decide?
On the other hand, if used as loose guideline, this could be helpful in making certain decisions — when considering how familiar the target audience is with certain concepts. In the example about the French Série Noire, it makes sense that the translators chose to use American-sounding names to preserve ‘authenticity.’ (Although the use of the word “deceive” might indicate otherwise.)
(I hope I am making sense and not just rambling…)
I appreciate Nida’s distinction between formal and functional equivalence. As a bilingual person, functional equivalence is so important — there is never a one-to-one mapping between ideas or words.
Interesting to think about whether translating a translation might actually work better when localizing websites. If you could translate from a language/culture that is closer to the one you want to translate to, you’d be likelier to get a better translation?
“[…] foreignism, calque, or a semantic expansion. Each of them changes the target language by one item, with possible repercussions over time on the use and form of other words. But cultural substitutions would simply put some other, more or less analogous activity current in the world of Aramaic speakers in the place of ‘jazzercising.’”
What could be some unintended repercussions of cultural substitutions, though? Isn’t it important to explain the context of the “fig leaf” to the Malay? Isn’t it important to preserve the original expression? I guess the intention/reason for translating plays an important role in answering those questions.
After reading further, Bellos does address the issue of preserving the original work. However, he makes the distinction between translating UP vs. DOWN. I’d argue it’s more about the purpose of the translation, rather than the direction. But yeah, in general, everything depends on context!
I am also curious about the power dynamic and Euro-centrism of translators (at least in the examples mentioned in these chapters). There are no examples about translating from, say, Hindi to Japanese. Those two cultures have a very different historical context than European countries and Japan or European countries and India. Maybe Bellos will address this later on… * continues reading *
It all comes down to “paying respect where respect is due.”
“[…] the reciprocal flow of translation between any two languages is never equal and in most cases unbalanced.”
Yay, Bellos acknowledged it!