Typed Japanese
github.com
github.com
and : Phrase \ Conjunction / Phrase
do : Subject \ Verb / Object
and : Phrase -> Phrase -> Conjunction
do : Subject -> Object -> Verb
Last time I looked at it they were adding polymorphism and a couple of people were starting to think about dependent types. It was mostly linguists interested in it but it was hardcore math and CS. Can't remember the names involved. Carl H. I forget what the H stands for. Hmm. Some fanous type theorists had written about them too.
Anyway, seemed relevant.
A good starting point is https://en.wikipedia.org/wiki/Combinatory_categorial_grammar
This is famously used in interpreting the status of "Socrates is mortal" ("Socrates" is the argument) and "Every man is mortal" ("is mortal" is the argument).
Unfortunately it didn't scale as well as opaque DNNs trained on huge amounts of data.
https://dic.pixiv.net/en/a/Why%20did%20we%20play%20Haruhikag...
Japanese is not unusual (on this particular point). That is probably the most common structure for a language. Swahili verbs are entirely regular and can have a long sequence of prefixes which follow a strict order. kula - to eat. usingalikula - (you) would not have eaten. halutayakula - (we) will not eat (those noun class 6 objects)
Human languages are, in a sense, "infinitary", so that they aren't simply a set of fixed phrases. However there are, depenging on the language, also patterns they don't allow. (And some patterns _any_ human languages don't allow.)
This logically necessities language to have regularities. We capture these regularities as "generative grammars".
For instance in English you'll need to know if a subject is a third person singular or not to adjust the verb. Or even just singular vs plural for the adjectives and pronouns. And you'll need the gender in gendered languages (French, Italian etc)
In general, Japanese doesn't require that kind of care: the verb doesn't change, adjectives don't either, most of the words are unaffected if you swap semantically different parts.
There is a suprising amount of subtlety here. What you need to known is if the subject is syntactically singular or not. Once you have that, it does not matter is the subject is actually singular. For instance, if you are talking to a single person, you say still say "you are" instead of "you is", because "you" is syntactically plural, regardless of its actual meaning.
This is very useful for generating grammatically correct sentences (or testing the grammar of existing sentences). The set of features that cause this type of inter-word agreement are called phi features (typically count, gender, person, and case). Since phi features are purely syntactic, you tag words with them just like you tag words with their part of speach.
In real speach, this can get weird, as it is possible for the same realized word to have different phi features depending on its sementic meaning. However, if all you care about is determining if a sentence is gramatical, you do not care if it means what was intended.
As an aside regarding Japanese, there have been some attempts to model Japanese politeness markers as a form of agreement; however, I have never found those arguements to be convincing.
It's a bit broader IMHO. For instance, plurals are handled as singular based on the semantics, even in written language. I liked how different authors handled the Olympics in wikipedia for instance
> The Summer Olympic Games, also known as the Summer Olympics or the Games of the Olympiad, is a major international multi-sport event
vs
> The modern Olympic Games (Olympics; French: Jeux olympiques)[a][1] are the world's leading international sporting events.
Same way a singular nouns will be handled as plurar if the speaker recognizes them as such ("The UN are planning a meeting")
There will be arguments on whether these sentences are grammatically correct, but I'd be more for the descriptive approach than the prescriptive one, as at the end of the day usage wins.
> attempts to model Japanese politeness markers
I kinda feel for them as it's the most evolving part of the language IMO. As people get used to a level of politeness, it loses it's status and a stronger form becomes needed. Basically there's a politeness inflation happening through the years, and it's all based on how people perceive it, so logic is also not an element of it.
This seems a little confused. If a language's verbs are entirely regular, the normal point of view would be that that language doesn't inflect verbs at all.
For example, an English verb can include up to four prefixes, which occur in an order that never changes and which do not depend on the verb. In this sense, that part of the grammar is "fully regular". (Note that all of these prefixes are verbs, and with the exception of the first set, they inflect like verbs. The first set inflect only for past tense.)
[Those prefixes, in order, indicate: modality (there are many verbs that can occupy this space, but they're all mutually exclusive with each other), perfect aspect (exactly one verb can occur here), continuous aspect (again, exactly one verb can occur here), passive voice (two verbs, mutually exclusive). Any of the sixteen theoretical combinations of these is allowed.]
But that's all periphrasis. English verbal inflection is a different phenomenon: each verb has five forms (except be, which is special and has more): plain (take), past (took), third-person present singular (takes), active participle (taking), and passive participle (taken). This is where a verb might be irregular or not. A regular verb's past form adds the suffix /d/ to the plain form, the active participle adds /ɪŋ/, the third-person present singular form adds /z/, and the passive participle is identical with the past form.
All verbs are regular in their active participle and only be is irregular in the third-person present singular.† But verbs might have arbitrary past forms (go / went) and while the passive participles are never arbitrary, there is a set of irregular forms following a common pattern that descends from an earlier stage of the language. (taken / been / gone / known / gotten / ...).
There is a double standard over what counts as irregularity in verbs. By the standard we use for Latin, English has a total of one irregular verb. A Latin verb may have a perfective stem that is arbitrarily different from its imperfective stem (compare ferre with tulisse), but this "doesn't count" - we say that it's necessary to memorize four forms of any Latin verb, and with English verbs having only five forms and two of them being always regular, it's a real stretch to find verbs that require you to memorize more than four forms.
† edit: this was wrong; have is also irregular in the present third-person singular.
It's hard to call them irregular as to the third-person present singular; there is a rule there and all of them follow it. They're just different from non-modal verbs.
You could fairly call shall and will irregular in their negative forms shan't and won't.
The passive marker get is irregular in failing to be an auxiliary verb, unlike all modals, both aspect markers, and the passive marker be.
And then there are a bunch of nuances that don't follow logic. You just need to learn about them. For example, word "you" is always considered informal, a bit intimate and a bit rude in spoken Japanese, it can be used in formal written Japanese.
Even informal Japanese comes down to basic patterns for grammar.
>For example, word ...
Word choice is not grammar. Knowing how to use words correctly is the hard part.
If you spoke English while only sticking to formal grammar rules, you'd be unable to communicate like a normal person at all.
I think the comment on the word 'you' sounding informal is a misunderstanding. People use a common formal noun for 'you', あなた, frequently; it's just the nature of Japanese emphasizes social relationships and makes it easy to drop references to people, so using such formal references can be stilted in the wrong context. But that's a matter of word choice, not grammar.
It's not really practical, but it is interesting to compare it structurally to programming languages.
A verb (or verbal adjective) basically forms the "root" of a clause, and typically appears at the end of it. If we related that to the concept of a function, it would be like writing them in postfix form, with whatever comes before it acting as the arguments to it.
(args)func
(WORDS)VERB
(arg1:subject, arg2:object)VERB
;eg
(NOUN:ga, NOUN:wo)VERB
(NOUN:wo, NOUN:ga)VERB
A verb clause which appears before a noun modifies the noun, which bears similarity to a subexpression.
(NOUN:wa, ((NOUN:ni)VERB NOUN):wo)VERB
(NOUN:ga)VERB ka
> Compared with a typical programming language, Japanese is structured the opposite way - from right to left.
How consistent is that across different types of phrases?
English is generally described as being strongly right-headed. But that's really a statement about English noun phrases. Prepositional phrases are left-headed. Determiner phrases, if that's your thing, are left-headed. Verbs are normally positioned between their subject and object, with other modifiers generally after the object.
I had a Chinese tutor who complained that everything in Chinese was the other way around compared to English, but as far as I can tell the odds are about 50-50 that any given structure will match or reverse. Mandarin sentences are SVO just like English. Mandarin prepositions go after their nouns instead of before. Mandarin adjectives go before their nouns. Mandarin noun compounds work the same way as English noun compounds. (牙刷 "tooth brush" is a brush, not a tooth.) Mandarin verbs have indirect objects before the verb instead of after. Mandarin discourse particles occur at the end of the sentence instead of the beginning.† Mandarin relative clauses work just like English noun phrases, which is funny because English relative clauses have their own bespoke ordering. (English: the man who has five sisters; Mandarin: has five sisters who [the] man.)
The big lesson I drew from this is "it doesn't make a lot of sense to describe languages overall as having a particular orientation". Is Japanese more consistent, or is orientation more sensitive to the specific structures you're using?
† Note that "beginning" and "end" aren't the only options. In classical Latin and Greek, particles like this always appear as the second word in a sentence.
The subject/object/means/destination/etc are flexible and can appear in any order - but typically subject appears before object - `NOUN ga NOUN wo VERB`, `NOUN ga NOUN ni VERB` (where `ga` denotes the subject), whereas `NOUN wo NOUN ga VERB` would be atypical, but not incorrect.
When a verb (or adjective/adverb) appears before a noun, as in `VERB NOUN`, it modifies the noun - acting like an adjective - `VERB NOUN` is like "NOUN that/which VERBS". All adjectives work this way - ADJ NOUN, because verbs and adjectives are not really disjoint word classes in Japanese. Even the noun adjectives ("na-adjectives") function this way - they appear as `NOUN na NOUN`, but `na` is secretly a verb disguised as a particle. Examples:
sora ga aoi (The sky is blue)
aoi sora (Blue sky) [i-adjective]
sora ga kirei desu (The sky is pretty)
kirei na sora (Pretty sky) [na-adjective]
sora ga haiiro ni naru (The sky turns grey)
haiiro ni naru sora (Sky which turns gray) [verb clause as adjective]
Verbs can appear out of order - for example at the start of a sentence - but the meaning is understood based on tone or pause in the speech - basically, if what were being spoken were to be written it would be with a comma after the verb. `VERB, NOUN`
Example:
watashi wa hara ga hetta (I'm hungry).
hara ga hetta, watashi (Hungry, I am).
hara ga heru watashi (Me who is hungry) [no pause or comma]There would be a better argument that this was an example of left-headedness if subjects also went after verbs.
Also, the morphology in verb inflection and case declination is often head-initial, if it makes sense to say that about morphology. Word formation seems to be kinda mixed.
But Japanese syntax is a lot more consistently head-final than with how many European languages have their parameter setting.
Grammar checking basically needs AI - you need to train some model to understand common phrases and sentence structure. Before LLMs there was software like MeCab[1] which done this, and gave good results, but modern LLMs are much more capable.
To give an easy example: 9つ (9 things) is natural, but 10つ sounds extremely strange. However, 10個 sounds fine. When the number is large enough, it's also common to not use 助数詞 at all.
Sometimes, grammatical mistakes are natural Japanese. For instance, there is a concept of ら抜き言葉 (words with ら dropped), where people will say e.g. 寝れない ("I can't sleep") instead of 寝られない. This is an error in conjugation, yet it's natural language and applies to a few other words, too.
Validating both grammar and word choice is still insufficient to judge naturality of a Japanese phrase. A common "mistake" made by many Japanese is writing 「違和感を感じる」. The verb is redundant because of the 「感」 in 「違和感」. The "correct" word to use in this case is 覚える. In practice, however, either choice of word is understandable and considered correct (except to those with the trivia of 「違和感は覚えるもの！」)
Sometimes, redundancy makes phrases considered incorrect (see 二重敬語 for an example). In other cases, nobody will debate the correctness of the phrase.
Besides, the grammar this project uses is not likely to reflect accurately the actual grammar of modern spoken or written Japanese, and it's likely not to be even nearly complete; that would mean it would also have a quite lot of false positive "ungrammaticals".
Something _like_ this can certainly be used as a part of a grammar checker. But in that case, you shouldn't implement it in TypesScript's type system in the first place.
Most of the "rules" are common patterns made into guidelines, and they'll change depending on the speaker, context, society of the time (the "correct" way is fully dictated by the majority). And you could break the grammar rules as long as the other accepted guidelines are OK.
As a parallel you can learn to mechanically drive a car, but driving it "correctly" will require full knowledge of traffic code, societal rules and how to reasonably handle conflicting situations, including crashing it into a tree if it means avoiding a packed school bus.