Re: [edict-jmdict] Re: POS tags - keiyoudoushi

[Jim Breen <jimbreen@*********>]

Hi Leonardo,

I don't think I disagree with what you're saying as I think we're both
heading in the same direction, perhaps describing it differently.

When I wrote "but you have to work very hard to use it as a noun", I
was thinking
of cases like 割高, which 大辞林 marks as "名 ・形動" but it's very hard to come
up with English translational equivalents which are nouns. None of the
JE, JF, etc.
dictionaries I looked at attempt to gloss it as a noun. Looking at the few valid
WWW usages I can see of 割高が, it seems to be meaning "state of being
fairly expensive".

I'm quite happy to tag it as (adj-na,n), with the "n" largely
reflecting what's in 大辞林.

I pretty much agree with your suggestions, but I'd add that they
should be influenced
by frequency of use. If a word in primarily ~な make the first POS
"adj-na" and gloss
it as an adjective. If it's more commonly ~が, tag and gloss it as a noun.

Cheers

Jim


On 12 June 2017 at 18:21, Leonardo Boiko leoboiko@namakajiri.net
[edict-jmdict] <edict-jmdict@yahoogroups.com> wrote:
>
>
>
> > There are cases where 大辞林 has "名 ・形動" but you have to work very hard to use it as a noun. I guess there's still no harm having an "n" in the POS, if only for historical reasons.
>
> What do you mean use it as a noun; in translation, or in Japanese? Of course, relying on establish dictionaries is good practice and a perfectly sufficient criterion, but when in doubt, I would argue for Japanese-centric, syntactic definitions of "noun" vs. "adjective", rather than translation-based or semantic.  That is, if the word non-markedly takes case particles (-no, -wo, -ga, -wa), then it's a Japanese noun. If it takes -na to qualify a noun it's a na-adj, and if it takes both it's both. This gives a more objective way to test using modern corpora, and bypasses issues with counter-intuitive cases like byōki (not an adjective even though 病気の人／彼が病気だ feels adjectivey, in translation at least).
>
> There's a number of pure na-adjs which normally don't take case particles (except if converted to nouns with -sa ), like shizuka.  Other words, like byōki, are pure nouns, which take case particles and don't take -na. I bet this test selects the same set of words marked as exclusively (形動) and (名) in Daijirin and others. Still other words are ambivalent, and may take both kinds of particles; those are, I think, the ones marked as both classes in Daijirin. (Once I tried to measure this, and found that usage proportions vary continuously; some words are used ~50/50 as nouns/na-adjs, some ~80/20 as one or the other, etc.)
>
> The ambivalent words can be further subdivided in two cases, and this time the distinction has to be semantic. Some words seem to change the nuance when used with case particles vs. with -na (平和の国 / 平和な国）. The other class doesn't seem to change meaning in any appreciable way (真黒の靴 / 真黒な靴). It could be argued that heiwa-words are underlyingly pairs of distinct, homophonous words; while makkuro-words are truly class-ambivalent single words.  In my personal opinion, FWIW, it would make sense to add distinct glosses for heiwa-words, and not to makkuro-words.  This seems to be Kenkyūsha's approach:
>
> > へいわ１【平和】 (heiwa)
> > peace; 〔和合〕 harmony.
> > 〜な peaceful; pacific; amicable; tranquil; harmonious.
>
> Vs.:
>
> > まっくろ【真っ黒】 (makkuro)
> > 〜な[の] deep-black; jet-[coal-]black; inky(-black); sooty; 《as》 black as coal [ebony, pitch, ink, soot]; filthy; 【動】 piceous.
>
> So my suggestion for editorial guidelines would be:
>
>  1. If the word is used with case particles / marked as 名詞 in references: tag as (n)
>  2. If the word is used with -na / marked as 形容動詞: tag as (adj-na)
>  3. If both 1 and 2 apply, and furthermore the meaning (in Japanese) changes between 1 and 2: gloss both meanings.
>
> In this way the information bit that "this word changes meaning (or doesn't) when you use the other particle" wouldn't be lost in JMdict data.  If you add POS tags for the glosses, like Kenkyūsha's heiwa example above, it could even allow dictionaries like Rikai to peek at the following particle and select the right subset of glosses.
>
>
>
> 




-- 
Jim Breen
Adjunct Snr Research Fellow, Japanese Studies Centre, Monash University