Stricter deinflection conditions

[Casheeew]

Add more conditions to transformations to avoid matching other words in the transformation process. (For example, prevents こうて from matching 斯くて)
also relocates the kansai-ben transformations so that verb transformations stay together

[codspeed-hq]

CodSpeed Performance Report

Merging #644 will not alter performance

_{Comparing Scrub1492:stricter-deinflections (b5b5cf1) with master (0e9c28f)}

Summary

✅ 2 untouched benchmarks

[github-actions]

✔️ No visual differences introduced by this PR.

View Playwright Report (note: open the "playwright-report" artifact)

[StefanVukovic99]

I think the way the past (ta) and adv (ku) rules are used here is correct, the suffixIn and the conditionsIn match. However, iru is used in the opposite way, conditionsOut matching suffixIn:

"name": "progressive or perfect",
"rules": [
    {"suffixIn": "ている", "suffixOut": "て", "conditionsIn": ["v1"], "conditionsOut": ["iru"]},
    ...
]

which makes it look out of place on -shimau because it doesn't match anything:

"name": "-shimau",
"rules": [
    {"suffixIn": "てしまう", "suffixOut": "て", "conditionsIn": ["v5"], "conditionsOut": ["iru"]},
    {"suffixIn": "でしまう", "suffixOut": "で", "conditionsIn": ["v5"], "conditionsOut": ["iru"]}
]

Should iru be renamed to te or conjunctive?

[Casheeew]

I have also thought of this, in addition, "iru" is being used for adverb くて form as well:

yomitan/ext/data/language/japanese-transforms.json

Line 345 in 376bac7

{"suffixIn": "くて", "suffixOut": "い", "conditionsIn": ["iru"], "conditionsOut": ["adj-i"]},

(which is obviously semantically incorrect)

I also dislike the term conjunctive as I think grammar in Japanese does not fit perfectly into the grammar model of Western languages, but at the same time I don't know enough about linguistics to have a formal opinion. Open to change.

[djahandarie]

I also dislike the term conjunctive as I think grammar in Japanese does not fit perfectly into the grammar model of Western languages, but at the same time I don't know enough about linguistics to have a formal opinion. Open to change.

Agree. I think いる performs too many distinct semantic functions to give it a name like that, I think just "iru" is fine.