3.8 Translation hacking
The first rule of translation hacking is under no circumstances should you touch a machine translator; they are certainly fascinating pieces of kit and they have their uses but for the foreseeable future they have no place in a creative work like translation hacking9 and especially not for nearly entirely unrelated languages like Japanese and European languages.
Translating games is a very popular activity among ROM hackers and if you go outside ROM hacking and ROM running circles fan translation/ROM translation will almost be a synonym for ROM hacking. Most translation is from Japanese to English, Japanese to Chinese or Japanese to another language of European origin (although Arabic, Russian, Korean and Thai translations are growing in number) and, unlike almost all the other areas of ROM hacking, this is not simply a matter of technical prowess as it requires language ability as well. There have been a few ROM hacker translators in the past but they are rare and usually were more one than the other so it usually means setting up a translation group to get things done. This certainly does not mean you as a ROM hacker would not do well to know a bit about the language you are translating from and this typically means Japanese.
Perhaps above all else you should recognise and appreciate that despite languages sharing many things in common (verbs, adjectives and nouns for one) they may also lack things (English does not really have gender for words, some languages have two and languages like Russian have three) and do things differently in things like adjective ordering. On the subject of word ordering most modern games that use markup/placeholders can work around it, or even change it without any negative consequences. If the thing to be replaced with a value is at a fixed point in a script you might have to work around it which can be troublesome or accept a slightly clunky translation. Some games have encoded larger pieces of text as part of the variable/markup though and that leads to things like innkeepers saying “it costs gold 600 to stay here for the evening”.
Equally languages are built upon hundreds of years of history, history which seriously informs how they work; Japanese culture was and in many ways still is strongly informed by notions of social hierarchy which influences the language no end and can be quite hard to translate/convey effectively if the culture of the language you are translating into, which is to say most European languages, does not have such a history. Likewise they have various levels of influence upon and from other languages; Chinese formed a large chunk of the Japanese written language for instance but nowadays they are far from entirely mutually intelligible (see kokuji and kokkun) and nowadays you are far more likely to encounter minor tweaks upon English words in Japanese than you might have been a hundred years ago.
The study of the differences between languages is seemingly quite a common thing for those in scientific and technology fields to engage in, however despite it being great to have an appreciation of it is not directly related to translation hacking so time to get back on topic.
The debate over whether a translation should be kept literal or be able to be adapted a bit has raged for years no end in sight, indeed several retranslation/cleanup projects have been set up to fix problems with official and unofficial translations over the years. Each and every side has valid arguments here with the only real consensus being changing a work drastically and calling it a translation is pushing the limit at bit (although if the end result works it can be quite interesting) and having your translation be readable is a good thing; note that readable and understandable only with a frame of reference for the franchise or a given culture/history is not necessarily the same thing, however such a thing can inform how you proceed with a translation.
Things one should appreciate about the Japanese language when playing ROM hacker (hacker side of the fence)
3.8.1 The types of Japanese characters and how they work -
Hiragana, Katakana and Kanji form the basis of the language
Kana. A collective term for the Hiragana and Katakana they are the basic constructs of the language with Katakana usually being used for loanwords/foreign words, also where the Hiragana are fairly freeform the Katakana are somewhat angular in appearance. Hiragana tends to be reserved for native words and both are phonetic, which is to say several are combined to make words.rikai.com shiftJIS has examples of both.
There are some fairly accepted ordering/sorting methods with Gojuon order being the most popular. Common ROM hacking table making tools like TaBuLar should add it and few games deviate here. On the flip side be aware the ordering might change, that standard tools might well leave out some of less common, and possibly obsolete, ones the script writer might use (or the font that was borrowed had included) or that the game might add entire characters for characters with punctuation (see Dakuten, Youon/Yoon and Handakuten although most good table creation tools should be able to add these) so be aware of this when constructing tables. ShiftJIS and EUC-JP use the order but will also include Dakuten for the respective characters between them and many games use or take their cues from the ShiftJIS and EUC-JP encodings.
Kanji Kanji are the elaborate symbols sourced from Chinese Hanzi characters (if presented with an unknown Asian language script one tends to tell Chinese and Japanese apart by looking for the simpler Kana which Chinese lacks) and there are a very large number of them. Indeed there is no upper limit, although a book known as Dai Kan-Wa Jiten is considered one of the more complete listings and almost a must for any translator. Even among native speakers they are considered the harder part of the language and many translators/language grades will reference various lists of them; most day to day language usually stops short of the 2000 mark. Unlike Kana they are ideograms and one symbol tends to represent an idea, although they can still be combined to form related concepts/compound words.
Importantly for ROM hacking there are various types of works and fields (medicine for example) that aim to reduce the number of them in their text. One of the more notable examples would be that shounen manga and anime do attempt to reduce the amount seen, a practice quite favourable for ROM translation as they tend to produce a lot of fairly interesting games that do not always make it outside Japan.
A few people have tried to make something resembling an ordering for Kanji and there are a few things you can learn that might help but in practice there is none and this means relative searching, which is very useful when making tables for Roman language games, is at best very tricky and more often largely ineffective.
If you do w want to try relative searching you will want to have an idea of the existing ordering such as from a font, via fiddling with the ram, via the name entry screen or something similar before resorting to a value search or an abstract search. It should also be noted that as far as encodings seen in the wild go there have been somewhat logical methods including order of appearance in the game text (or text from another game if the table is borrowed), character count in the text (most popular first for example), ordering borrowed from a known encoding (even if the encoding values are different) and similar things so do consider that as well. Also as game development/publishing companies may not be wanting to bother defining a new encoding for every game they make they may well share them between games, either in whole or in part, or use existing encodings to create new ones.
As mentioned above there are lists of Kanji you might want to look at to help make tables, it is far from sure fire though and might be more useful to those doing the translation; after all modern games are usually written by modern writers for modern audiences, both of which tend to be products of modern education and modern language reform. One should be wary though as games do like to use older symbols as magic symbols and decorations, to say nothing of wordplay and language appreciation being a fairly common thing in Japanese culture.
Historically there were things known as moji (categories) which also presumably led to the slang word “moji bake” for when the encoding used by the decoder is wrong (not to be confused with cavespeak which is where a 16 bit encoding is swapped for an 8 bit one which breaks the untranslated text), these might be useful to know about but this is getting into Japanese orthographical history and even with a decent knowledge of Japanese will probably not be that helpful for those tasked with reverse engineering game text files. On the other hand you should know what the terms radical and stroke ordering mean as they are in very common usage. To that end radical refers to the base stroke/component the Kanji in question started at/as and stroke is quite literally how many brush/pen strokes are needed to complete the Kanji in question. It is not a true ordering system but when looking things up these will be what is used in the vast majority of cases.
ROM hacking tools, or even general use tools, to add vast lists of Kanji in a manner similar to TaBuLar’s abilities in Kana and Roman languages do not really exist although TaBuLar will support adding lists of 16 bit characters if you have them premade. Should you have a common ordering (or a couple of fragments of it they can arrange and add to accordingly) or if pulling it directly from the game; OCR (optical character recognition) for Japanese does exist and is even in a couple of ROM hacking tools like crystaltile2, although it works better on Gothic fonts and is troubled by some of the more fanciful/handwriting esque fonts like those in the Kaisho and Gyosho lines and sometimes even Mincho, also if you get a tool that can do it Tensho and Reisho, which appear very different to other Kanji, is not that bad to OCR. In the end you do have some options but do expect a high amount of manual work until proven otherwise.
As far as basic regular expression type searches are concerned Kanji and the kana are able to be anywhere in the text and nowhere, naturally if you improve your Japanese knowledge you will there are situations that commonly call for one or the other, or a set thing (think u always follows q in English), and you might get something done.
All that said if you need to change a font as part of the game hack and keep enough characters/language for your translators to still do some ongoing play testing then Kanji are usually the better choice for something to lose.
Kanji ordering Although it was stated there is no order there are groupings which you might wish to learn to recognise or at least appreciate. The most notable versions of this are the
- ?-moji. Although now considered if not obsolete then not best practice by many, both in Japan and even outside it, the idea of categorising kanji into groups with the suffix moji it is still known of and influences things. Six groups exist, four of which go by the basic stroke/aesthetic and the latter two by their meaning. Lists are hard to come by and not that widespread but so you know if your translator tells you it is grouped roughly like this you will have at least heard the term before.
- kyoiku. A list of kanji as specified by the Japanese ministry of education and broken down by school year/age range. Depending upon the person you are speaking to the complete collection/concept of the grade by grade breakdown can go by the name gakushu although that is not ideal.
- joyo. This would be the complete list (kyoiku plus the ones from secondary school age ranges) of common use kanji.
Once again various game specific methods have been seen and many custom methods will share portions or even whole sections with shiftJIS, eucJP or something similar as it is just as awkward for the original programmers to define a new order as it is to do it yourself. Equally there are examples of games using an ordering based upon the order in which characters first appear in the script and orders based upon how common characters are.
Furigana Furigana is supposed to be to help with pronunciation for Kanji (aimed mainly at younger people and those learning the language) but has been seen to hold little jokes, extra explanations and much more, basically turning it into the equivalent of an asterisk or a footnote/reference. It was not that common on older systems, mainly for technical reasons, but the DS sporting a nice touch screen and enough resources that you do not always have to heavily optimise things has seen it used extensively. It usually comes as a type of markup or flagged text but not always, pointers for example could handle it, and you can use the space they used to help you fit more into the file/memory if you are limited there.
Others (yakumono) Although Japanese has many of the same concepts as languages using the Roman character set in terms of punctuation, shorthand and constructions stemming from them the characters they use can be radically different. sljfaq has more on this. This can allow for some hex analysis type techniques to return.
3.8.2 Japanese glyphs/characters and observations on the language
There was some mention of the types of fonts seen in Japanese but in general the characters are fixed width and stay within the lines. This contrasts with the Roman alphabet which has things like ijlt, and on the flip side WQMK, as well as punctuation which almost always sits right next to the previous character without much of a gap, and on the matter of lines thing characters like j y p q and f in some fonts/writing styles.
Also note Japanese does not have space characters per se with any you see being largely aesthetic or for line wrapping purposes; you will probably encounter this when translating a game from Japanese that has a text engine built by a Japanese programmer for use with the Japanese language but the same could be said of most aspects of translation.
Tategaki In short Japanese is usually written horizontally, much like English and European languages, and it is even read left to right. However it can be written vertically and read top to bottom with the name for the concept being tategaki. It is usually only reserved for introduction sequences and artwork as far as games and most modern Japanese goes so it does not tend to trouble translation, however it was seen in Sigma Harmonics on the DS.
Equally although it is not strictly legal in English grammar the use of it for short amounts of text (for instance how often have you see a hotel, cinema and some such use it in a sign) or simply to cause a mental association with Japan; it is a fairly recognised visual effect and given it will also probably be floating over an anime style image or ink wash painting it helps too. To that end you might wish to leave it.
Romaji One of the main reasons Japanese is considered hard to learn is that unlike learning most other languages you also have to learn a new character set. To lessen this and in some cases allow Japanese to be written without a proper input method editor (IME)/Japanese keyboard (which remember could be thousands of keys in size if you wanted to include even a small subset of kanji) ways of overlaying the Japanese language to the Roman alphabet were devised. There are several with the most popular, although not necessarily most liked, being Hepburn. Very few Japanese games will use it and occasions where it might be instead often opting for all English and almost definitely Arabic numbers but it is a fairly important concept and doubly so when dealing with multiple languages at once.
Sentence length Whether due to the existence of kanji or maybe just in general the written version of the Japanese language is quite often shorter than the English translation which ROM hackers and the official translation teams both run up against. Speech is a different matter and is usually roughly equivalent in length which is quite nice for the would be translation team usually as it will be peppered with honorifics and such things.
Wordplay All languages and humour using them have many examples of wordplay but Japanese is especially noted for it and indeed many translators have missed things like this and delivered a translation very different in tone to the original over the years because of it. With a few very notable exceptions (Rudra no Hihou on the SNES usually being first in that list) not many games have used language play as a core concept/mechanic similar to how English gets treated when magic gets brought up.
It is also yet another reason why machine translation is not to be trusted.
3.8.3 On language
Translating a game that already has a translation out there has happened several times over the course of things and this is usually because the translation does not fit with various ideas; it could be an awful conversion and full of bad slang, it could be that the translation team lacked technical skills (even with the source code some games can be hard to translate11 ) or it could be that the translation team were hampered by the developer/publisher and had to change the tone of the language or word choice to appeal to a broader audience or meet censorship requirements. Although this is a very interesting field in ROM hacking, and depending upon what you are doing this can even be a solo project, it is not what this section is about.
Several discussions have been had over the years on the matter of how to translate certain concepts. It also applies to Japanese with a good example being found in many Final Fantasy and other Square Enix titles where the folklore and literature of Europe and the Middle East is often a source for monsters and themes, not to mention other games use Chinese history and folklore as a theme. Much like the translations of games from Japanese to English the translations of items and concepts to Japanese can also see people butcher a meaning of something.
On a different note Japanese, English and other languages are often quite different, there are also loanwords that might have had meaning lost at points. For instance you might know Sushi is not raw fish but more than a few people still consider it as such, not to mention where it is a normal enough food in Japan it is still quite exotic elsewhere and as such you might consider losing sushi entirely and replacing it with a common foodstuff of your intended location if the setting is a contemporary city or something.
3.8.4 Right to left languages and translation.
It has been mentioned a few times already but it is worth having a quick section here as well. Not all languages favour reading from left to right and this can trouble translation hacking, especially as more and more countries that use such writing systems start to make games and seek those made before. There is tategaki mentioned in Japanese but most will be more concerned with Arabic and Hebrew, as well as the languages that took one of those as a base and ran with it. Fortunately languages that alternate direction, a concept known as boustrophedon though strictly that should probably also mirror the line, are not in common use. Back on topic those that live in locations, want to learn languages from said locations and otherwise are interested are many in number and games are fairly universally enjoyed by people. Typically there seem to be three schools of thought
- Just leave it left to right
- Cheat
- Full hacking
Owing to most computer and standards development tending to happen outside of the locations that use RTL written languages most computing that supports it has not quite filtered down into the consoles yet, for a related concept see 8bit and 16 bit encodings and the historical fun and games there. Even on computers though things are not great so a lot of people get used to reading things on a screen differently to how they might read a piece of handwritten or printed text. Indeed RTL and such computer programming is a relatively new concept. To that end as long as the glyphs are there and in the right order it works, you probably do something similar when you read 2^4 as 24 on a computer. Not ideal by any means but perhaps better than nothing.
Cheating then… if you have ever watched someone new to text editing use a text editor then you might have seen them put text in the middle of the screen using space or tab and force new pages by pressing return enough times. A similar technique is used here and for every line of text enough spaces/tabs will be added that it appears on the right hand side of the screen. Usually here ROMs with the largest scripts will be chosen as the base ROM, even if they are not changing that language (those tackling the the N64 Zelda games tending to use the French ROM but the English script because of this). Again this is far from ideal but getting better.
Full hacking then. A surprising amount of the text engine, and possibly even the underlying hardware as far as references for screen position, will be predicated upon text appearing from left to right, however such a thing potentially not even as bad as hacking in a variable width font. The idea being where a text engine might do something like “draw character, add 16, draw another character, if end of line then start new one…” you instead tell it to start at the end draw a character, take 16, draw a character and handle new lines accordingly”. You will have to be wary about negatives and how it might handle going over (lose a character space and say if it is below the width of the largest character then new line is fairly safe here) but all in all it should probably be easier than a full VFW hack, which you may also have to do.
It should also be noted that various characters might be joined in up written Arabic, many will take the “it is a computer therefore it does not matter” approach but if you are being flash then perhaps consider it.
Some have said the invention of realistic virtual sex will spell the end of the human race but others would argue the invention of a machine that can do humour would be more damning and the first step to that would be being able to translate humour and/or wordplay from one language to another.↩︎
The traditional example is to provide a translation of a piece of a text and back again so using the paragraph just written from English to Japanese and back again: The first rule of hacking your translation should be touching the machine translation is located under any circumstances; they are attractive part of kit indeed, although they have their uses, for the time being, they are hacking translations for language completely unrelated nearly as well as the language of Japan and Europe, not particularly There is no place for creative work like.↩︎
A Japanese to Dutch translator over at Loekalization details many of the issues faced by those translating games “for real”. Some things are very much focused on professional translation but a lot of it also rings true for ROM hacking and even the professional stuff hints at the issues that are seen when it comes to ROM hacking.↩︎