2

My question essentially boils down to the following: Is it always possible to assign a (regular / irregular) reading to a single kanji within some word? It is related to this post: Kanji that don't use their specified readings:

The context is somewhat technical: I'm writing software that should map a kanji to its reading within a certain expression. I'd like to collect all irregular readings for the kanji and want to know, which irregularities are most common. I'm using a database of all official on/kun-readings and Jim Breen's dictionary and for this: For each of its ~190.000 entries, I try to match the native (kanji) expression character-wise to its reading defined in the dictionary. I reuse all irregular readings that could be uniquely determined (i.e. if a single kanji remains unresolved in an expression, I use the remaining reading part as irregular reading for this kanji. E.g. お兄さん -> 兄's reading in this case is にい). I'm not sure, if this is always correct though, here are some examples:

今日 -> きょう (is 今[きょ]日[う] correct?)

明日 -> あした (is 明[あ]日[した]correct?)

越幾斯 -> えきす (is 越[え]幾[き]斯[す] correct?)

Around 300 (exotic?) entries remain, where it is not possible to uniquely determine the reading of each kanji. How about the following exemplary ones:

七週の祭 -> ななまわりのいわい

海月 -> くらげ

Do these ones have to be learned as a compound? Thanks in advance for your help!

3

The short answer is no, it's not possible.
While most kanji can be mapped to their most common readings, as you have already shown in the question, there are exceptions where it's not so simple.
今日 -> 今{きょ} 日{う}
明日 -> 明{あ} 日{した}

But even in this case they can be mapped arbitrarily, right? Because it is arbitrary, asking whether it is correct or not becomes moot. Note, however, that there is no other way to divide きょう than into きょ and う since き and ょう is just nonsensical (I'm assuming you don't want a "null" kanji). So while these can be mapped like you have them, any mapping is going to be arbitrary.

You have also already mentioned 熟字訓{じゅくじくん} like

海月{くらげ}

Again, there is no "correct" way to map these because these kanji don't have those readings, irregular or not.

The real reason I said it is not possible though, is the existence of 3-kanji 熟字訓{じゅくじくん} with 2-kana readings.
As in

大口魚{たら}
桃花鳥{とき}
百舌鳥{もず}
五十日{いか}

Granted this mostly occurs with last names, place names, and (usually) alternate ways of writing animal names.

Plus, you have to figure out what to do with words where the same kanji compound has 2 different readings (and meanings) depending on context.
As in

人気{にんき}   人気{ひとけ}

And then you have words where the order of the kanji is opposite the reading
As in

不忍{しのばず}

It's the first kanji 不 that represents the last kana ず.

This word also leads to the last category of readings problematic to mapping I can think of, hidden kana. Most often の, but there are others.
As in

不忍池{しのばずのいけ}

It's not possible to map the second の anywhere.

I think you could go a long way with just the regular/irregular readings, so for practical purposes it may actually be possible. But, even some of the most common words will give you many exceptions you would have to handle on a case-by-case basis.

  • These are 熟字訓 (assigning meaningful characters ignoring readings), not 当て字 (assigning characters ignoring meanings). 当て字 is something like 仏蘭西. – naruto Jul 16 at 16:01
  • Thank you and edited. – By137 Jul 16 at 16:06

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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