I'm casually ideating on a new orthography for Japanese that does not require Kanji.
Because it's hard to remember how to handwrite complex kanji (many people have character amnesia in real life due to smartphone usage), I casually wondered if it was possible create a Japanese orthography that was: (1) easily scannable (which rules out hiragana); (2) disambiguated words without Kanji; (3) still relatively compact? (which hiragana is not).
I figured a good substrate to start from was romaji + a new emoji system.
You know how people think LLMs can't invent things? o3 just invented this system whose goal is to maximally disambiguates homophonic Japanese words (performing the same semantic compression role that Kanji does today). This is the first iteration. After each romaji noun, it tags it with a geometric shape. These are the 30 tags o3 came up with (based on homophones requiring disambiguation):
| Diacritic | ◐ living | ▢ built / object | ⬢ nature | ◊ abstract | ⟐ action / event |
| ------------- | ---------------------- | ---------------------- | ----------------------- | ---------------------------- | --------------------------- |
| • top dot | ◐• adult / main person | ▢• building / place | ⬢• plant / flora | ◊• idea / thought | ⟐• movement / travel |
| – bottom bar | ◐– child / minor | ▢– tool / instrument | ⬢– water / liquid | ◊– emotion / feeling | ⟐– communication / speech |
| + right plus | ◐+ group / collective | ▢+ vehicle / transport | ⬢+ weather / sky | ◊+ social tie / relationship | ⟐+ creation / production |
| × left cross | ×◐ animal (non-human) | ×▢ food / consumable | ×⬢ mineral / material | ×◊ value / moral | ×⟐ perception / sense |
| | center bar | ◐| deity / honorific | ▢| document / media | ⬢| terrain / landscape | ◊| state / condition | ⟐| change / transformation |
| ‿ bottom arc | ◐‿ anatomy / body part | ▢‿ container / vessel | ⬢‿ energy / fire | ◊‿ knowledge / data | ⟐‿ rest / passive state |
And told o3 to rewrite it in the new system it just created:
gendai◊| ni okeru nihongo◊‿ no ippan-teki na hyōkihō◊‿ wa kanji▢| kana▢| majiri bun▢| de ari, kanji▢| to hiragana▢| (mukashi no hōrei▢| nado de wa katakana▢|) o majete hyōki◊‿ suru. kanji▢| wa jisshitsu-teki na imi◊• o arawasu go◊‿ ni tsukaware, hiragana▢| wa omo ni katsuyō-gobi◊‿ ya joshi◊‿ ni tsukawareru. chōsengo◊‿ nado de wa kanji▢| ga shutoshite ji-on go◊‿ ni shika tsukawarenai no ni taishi, nihongo◊‿ de wa wago◊‿ ni mo tsukaware, gairaigo◊‿ o nozoite hotondo no go◊‿ ni tsukau koto◊‿ ga dekiru. tabako⬢• (tabako) ya kappa▢– (kappa), kōhī×▢ (kōhī) nado dai-kōkai-jidai◊| ikō ni Yōroppa⬢| kara haitta goi◊‿ ni wa, gairaigo◊‿ de aru ni mo kakawarazu kanji▢| ga tsukawareru mono◊‿ ga aru.
It's pretty readable, and takes care of the homophone ambiguities (remaining ambiguities can be resolved through context). It also naturally resolves onyomi and kunyomi. Add italics and punctuation, and katakana is replaced.
(of course, it's incorrect in parts... because LLM)
But the idea has legs. It will probably not replace Kanji due to cultural inertia, but this could a kind of shorthand especially for handwriting.
I'm pretty impressed! o3 just invented something new. It combined romaji and a tag system that it hallucinated to design a new Japanese orthography. As far as I can tell (I could be wrong), something of this nature has not been done before.
Because it's hard to remember how to handwrite complex kanji (many people have character amnesia in real life due to smartphone usage), I casually wondered if it was possible create a Japanese orthography that was: (1) easily scannable (which rules out hiragana); (2) disambiguated words without Kanji; (3) still relatively compact? (which hiragana is not).
I figured a good substrate to start from was romaji + a new emoji system.
You know how people think LLMs can't invent things? o3 just invented this system whose goal is to maximally disambiguates homophonic Japanese words (performing the same semantic compression role that Kanji does today). This is the first iteration. After each romaji noun, it tags it with a geometric shape. These are the 30 tags o3 came up with (based on homophones requiring disambiguation):
I gave it this wikipedia JA text: And told o3 to rewrite it in the new system it just created: It's pretty readable, and takes care of the homophone ambiguities (remaining ambiguities can be resolved through context). It also naturally resolves onyomi and kunyomi. Add italics and punctuation, and katakana is replaced.(of course, it's incorrect in parts... because LLM)
But the idea has legs. It will probably not replace Kanji due to cultural inertia, but this could a kind of shorthand especially for handwriting.
I'm pretty impressed! o3 just invented something new. It combined romaji and a tag system that it hallucinated to design a new Japanese orthography. As far as I can tell (I could be wrong), something of this nature has not been done before.