Wondering if the translation of languages such as Chinese and Japanese into Romanized versions is accurate. That is, it doesn't lose information. For example, in Romanization of Chinese, they say there are variations in pronunciation. So what I imagine is when it is Romanized, you can't necessarily go back and translate it back to the original Chinese. Wondering if that is correct. If so, wondering how much information you generally lose, and where you loose it. If not, well that's great!

Basically wondering if the sound translation of non-sound-based languages (e.g. logographic ones for example) is accurate, or how much information is lost (preventing reverse translation).

  • Related question: linguistics.stackexchange.com/questions/20459/… Aug 29, 2018 at 15:26
  • 1
    Information is lost in Chinese and Japanese, because even assuming the romanization preserves the sound with full accuracy, and even accounting for context being able to help a reader reconstruct the original spelling, sometimes there are multiple logograms for the same word or portion of word (as in the same meaning, not just the same pronunciation), and the choice of which one gets used is stylistic and not predictable. In Japanese, it can also be a stylistic choice for some words where to write them in kanji or kana.
    – LjL
    Aug 29, 2018 at 18:48
  • While No Language is Fully Accurate before or after "Romanization". "欲穷千里目" cannot be Accurate without extra remakes. But you can still have the full information with "Yu-Qiong-Qian-Li-Mu" with the remarks.
    – PdotWang
    Oct 15, 2023 at 12:33

2 Answers 2


The purpose of a good romanization is to represent the phonemic distinctions of the language accurately, so that a native speaker who understands the romanization can get the same information from the romanized words as they would from those same words spoken aloud.

For example, if English were just being romanized now (from some other writing system), the romanization would ideally not have any silent E's, double consonants, or other irrelevancies: it would represent the pronunciation accurately, so that a native English-speaker would be able to pronounce an unfamiliar word accurately just from its romanized form. (While Shavian isn't a romanization, it's a good example of how these principles would be applied.)

And indeed, such a romanization would lose information! For example, "your" and "you're" would be written the same in such a system. But the key is, it's clear that English-speakers can always tell the difference from context: after all, spoken English doesn't make any distinction between those words, and we can understand spoken English perfectly fine.

On the flipside, such a romanization wouldn't necessarily represent all the details of the pronunciation in the way an IPA transcription would. For example, it might represent the vowels in "trap" and "bath" with the same symbol, even though some speakers pronounce them differently. Or it might represent the /k/ in "kit" and "skit" with the same symbol, even though there's an objective difference in pronunciation. The key is that these differences aren't seen as phonemic: they're predictable from the context and don't need to be represented explicitly.

(Many modern romanizations aren't actually perfectly phonemic; it's a balance. But that would be the ideal.)

So while a good romanization can indeed lose information, and can indeed fail to represent all the details of pronunciation, the key is that it contains the same information as the spoken language. Since languages tend to be spoken first and foremost, with writing as a secondary concern, this is generally considered the ideal.

  • No linguistic information is lost in spelling your and you're the same. They're pronounced the same in English and we never get confused about them in speech, only writing; hence they're not phonemic and shouldn't be part of a well-regulated Romanization.
    – jlawler
    Aug 29, 2018 at 15:54
  • Hmmm, so if the difference betweeen "rour" and "you're" isn't linguistic, what is it?
    – user6726
    Aug 29, 2018 at 16:01
  • 1
    @jlawler Indeed, that's what I tried to convey; how would you recommend putting it differently? (There's information lost, but it's not information that's needed, as evidenced by the fact that we pronounce both the same.)
    – Draconis
    Aug 29, 2018 at 16:46
  • @Draconis FWIW, I understood that was what you meant in your answer without any doubts. I don't necessarily agree that creating writing systems that are "purely phonemic" is "ideal", though, and personally, I think there are good arguments for preserving a distinction such as "your" being a word and "you're" being two words with an elision. Few "real" writing systems of today just aim to be a "sound recording" (even at a phonemic level), either: why do we have spaces and punctuation (and no, commas don't just map to pauses in speech)? Why don't we just write everything as IPA? I would not.
    – LjL
    Aug 29, 2018 at 18:45
  • @LjL Ah, I should clarify: I don't think that's necessarily ideal for an orthography. I think it's close to ideal for a modern romanization.
    – Draconis
    Aug 29, 2018 at 19:41

A theoretical issue arises in Japanese, in that 箸 'chopsticks' and 橋 'bridge' are Romanized as hashi but have different pronunciations (the accent is on a different syllable), likewise 今 'now' and 居間 'living room' = ima; 牡蠣 'oyster', 垣 'fence', 柿 'persimmon'; 鮭 'salmon' = kaki, 酒 'alcohol' = sake. It may also be necessary to put the words in context to get a difference in pronunciation, putatively in the case of 'chopsticks' vs. 'bridge': this depends on dialect as well. Writing systems typically omit tone diacritics (but may include consonant work-arounds as in the Romanized Popular Alphabet for Hmong or the Chao Romanization). The situation as I understand it is that tones are not normally marked in Pinyin except in language teaching materials (whereas my Japanese textbooks don't bother with any accent marking).

The extent of accentual significance in Japanese is fairly minimal, and normally people can distinguish persimmons from oysters in the grocery store, so signage could probably be all Romanized. Such information about word choice as is not conveyed in the spelling can usually be reconstructed heuristically. Chinese without tones is more challenging. The ultimate test of information loss is that comprehension problems arise in one writing system that are avoided in another. Since Pinyin has not replaced Chinese characters, I doubt that the question can be put to the test.

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.