gwalla | ローマ字 (Reply)

The existing systems of romanizing Japanese are unsatisfactory: Hepburn doesn't map directly to kana, Kunrei-shiki and Nihon-shiki's pronunciation is not obvious to non-natives (the primary users of rōmaji), waapuro is an anything-goes mess, and JSL is just odd. So, I've come up with my own solution, that I believe incorporates the best aspects of all of them. The advantages:

Pronunciation is unambiguous. There is only one case where the pronunciation the spelling suggests to a naïve speaker isn't really correct, but it's rare and close enough to be minor.
There is a direct mapping to and from kana.
Pitch accent—which can be phonologically salient—is marked, like JSL but unlike all other systems (including kana).
Can be written entirely in the Latin-1 character set.
A standard way of spelling a trailing small tsu.
Marking for rendaku

The latter two are not found in any of the other systems.

Of course, there are disadvantages:

That one case where the spelling doesn't precisely suggest the proper pronunciation.
Since it differentiates between homophonous kana, the spelling can not always be unambiguously determined from pronunciation by a naïve listener; however, the simpler spelling is usually the correct one.
Unlike NS and KS, it is not strictly "one consonant, one letter". I'm not sure if that's really much of a disadvantage at all, frankly.
Some verb conjugations seem slightly irregular. It shares this problem with Hepburn, but it seems to be a necessary compromise for good naïve pronunciation. And since kana are usually taught in organized rows and columns, I don't think it's much of a stumbling block.

So, without further ado, the rules of my system (hiragana is used for examples, but the rules are the same for equivalent katakana):

In general, spelling of individual kana is as in Hepburn
Geminate ch is spelled cch instead of tch (this is already common wapuro practice)
づ and ぢ are spelled dzu and dji, respectively, to differentiate them from ず (ju) and じ (ji). Dzu is the non-obvious bit (it's pronounced identically to zu)
Long vowels spelled as doubled vowels in hiragana (おお, for example) or with a vowel extender in katakana are spelled as double vowels, except for long o, which is spelled oh, and long e, which is spelled ey. For example, 十 (とお) is spelled toh instead of too.
An o-column kana followed by う is spelled with ou. An e-column kana followed by い is spelled with ei.
Pitch accent is marked with a diacritic on the accented vowel (this is why long vowels aren't marked with macrons; it's also why syllable breaks aren't makred with apostrophes, which resemble acutes and graves):

Accent on the first syllable (starts high, drops for the second syllable) is marked with a grave.

Mid-word accent (pitch rises until the next syllable, which is the peak) is marked with an acute.
e.g. 酒 sake vs. 鮭 sàke, 今 ìma vs. 居間 imá

Explicit syllable boundaries are marked with a middle dot (·), which is

allowed at any syllable boundary.
prohibited at any position that isn't a syllable boundary.
required after n, h, or y if the letter could otherwise be misinterpreted as beginning the next syllable (e.g. れんあい is ren·ai, rather than renai, which would mean れない).
required for syllable boundaries between vowels that would otherwise be treated as a dipthong (e.g. 思う (おもう) would be spelled omo·u, not omou).
recommended for syllable boundaries in an inflected word that are obvious but correspond to syllable boundaries that must be marked in the uninflected form (e.g. the past polite form of omo·u can be spelled omoimashita but should be spelled omo·imashita).
recommended for syllable boundaries that are unambiguous but not immediately obvious.
discouraged for syllable boundaries that are obvious.
If a middle dot is not available, a hyphen may be substituted, but this is less than ideal, and must be consistent throughout a text.

Explicit morpheme boundaries may be marked with a hyphen

Distinct words in apposition, or in a compound verb
Basically anywhere they're normally used currently.

A small tsu at the end of a kanji reading, indicating that the initial consonant of the following kanji in a compound is to be doubled, is spelled with a q. E.g. 一 has a reading iq, which is found in 一本 (ippon), 一般的 (ippanteki), etc.
Optionally, rendaku voicing may be indicated by surrounding the voiced consonant in parentheses: e.g. te(g)ami. This is an aid to someone who may want to look it up in a kanji dictionary: it tells them that the kanji is listed under the unvoiced equivalent (in this case, it's kami, meaning paper).