Extend support for irregular inflections of suru-verbs

[enellis]

Description

This adds support for irregular inflections of する-verbs by introducing a new inflection reason: 'irregular'.
This is used to handle cases where the verbs deviate from the standard する conjugation pattern.

The following irregular forms are now supported:

• ⚪︎せさせる, ⚪︎せられる (irregular causative and passive forms)
• ⚪︎しさせる, ⚪︎しられる (alternative causative and passive forms)
• 五段化-verbs (する-verbs inflected as す-Godan verbs)
• ずる-verbs

Caveats

• When handling irregular forms like 罰せられる, the structure could be misinterpreted as irregular < potential < potential or passive instead of the correct irregular < potential or passive. To prevent this, we now explicitly avoid reason chains such as potential < potential or passive or potential < causative, which would be grammatically incorrect anyway.
• Deinflection of vs-c class verbs is reverted, as these were only used in classical Japanese.
  Since we currently focus on modern Japanese inflections, retaining these rules could lead to confusing matches in modern contexts.

---

Fixes #1467. See the related discussion for more details.

[enellis]

@SaltfishAmi Could you help me with a good Chinese translation for irregular / 不規則 in the context of irregular verb conjugations?

[birtles]

Thanks for doing this!

• Deinflection of vs-c class verbs is reverted, as these were only used in classical Japanese.
  Since we currently focus on modern Japanese inflections, retaining these rules could lead to confusing matches in modern contexts.

I guess that makes sense. I don't suppose you have any examples where it gets confused?

Also, I'm curious about the vz class. It looks like we didn't handle that before (since we only matched vs-*). Do we have any example words there? I guess it would be nice to add a unit test or two for vz and some of the other cases handled by this (especially cases that trigger the ruleReasons.add(Reason.PotentialOrPassive); ruleReasons.add(Reason.Causative); block).

What do you think?

[enellis]

Thank you for the review!

---

• Deinflection of vs-c class verbs is reverted, as these were only used in classical Japanese.
  Since we currently focus on modern Japanese inflections, retaining these rules could lead to confusing matches in modern contexts.

I guess that makes sense. I don't suppose you have any examples where it gets confused?

What I had in mind are cases where both a す and する entry exist, such as 会す [えす] and 会する [かいする]. When matching an inflected form, you would consistently get both results, creating unnecessary noise.
This becomes especially problematic with forms like した, して(...) and しろ, which are not correct because classical Japanese used different auxiliary verbs to express those.
In my opinion, if we wanted to deinflect to classical verbs, we should do it properly and introduce additional WordTypes and add rules with the correct auxiliary verbs.

---

Also, I'm curious about the vz class. It looks like we didn't handle that before (since we only matched vs-*). Do we have any example words there?

The vz class refers to zuru verbs, which are basically rendaku-ed suru verbs. Modern じる-verbs, such as 感じる, evolved from suru verbs (感する) to zuru verbs (感ずる), and eventually to ichidan verbs (感じる). Some forms of zuru verbs, like 感ぜられる, are still used in modern Japanese, often to evoke a classical style. Maybe, from a modern perspective, we can think of ずる-verbs as an alternative inflection pattern for じる-verbs.

---

I guess it would be nice to add a unit test or two for vz and some of the other cases handled by this (especially cases that trigger the ruleReasons.add(Reason.PotentialOrPassive); ruleReasons.add(Reason.Causative); block).

Zuru verbs are basically just special class suru verbs, so they are included in the test cases for these. There is a test for every form of vz class verbs that is now being deinflected under the // ずる comment.:

      // ずる
      ['信ずる', '信ぜぬ', [Reason.Irregular, Reason.Negative]],
      ['信ずる', '信ぜず', [Reason.Irregular, Reason.Zu]],
      ['信ずる', '信ぜさせる', [Reason.Irregular, Reason.Causative]],
      ['信ずる', '信ぜられる', [Reason.Irregular, Reason.PotentialOrPassive]],
      ['信ずる', '信ずれば', [Reason.Irregular, Reason.Ba]],
      ['信ずる', '信ぜよ', [Reason.Irregular, Reason.Imperative]],

The block is tested by the following two cases, which are the reason why the block was introduced in the first place:

      ['発する', '発せさせる', [Reason.Irregular, Reason.Causative]],
      ['発する', '発せられる', [Reason.Irregular, Reason.PotentialOrPassive]],

If the block wasn’t functioning correctly, the reason chain would always include an additional Reason.Potential from the

  ['せる', 'する', Type.IchidanVerb, Type.SpecialSuruVerb, [Reason.Irregular, Reason.Potential]],

rule, causing the test to fail. Should this be made more explicit?

[birtles]

What I had in mind are cases where both a す and する entry exist, such as 会す [えす] and 会する [かいする]. When matching an inflected form, you would consistently get both results, creating unnecessary noise.
This becomes especially problematic with forms like した, して(...) and しろ, which are not correct because classical Japanese used different auxiliary verbs to express those.
In my opinion, if we wanted to deinflect to classical verbs, we should do it properly and introduce additional WordTypes and add rules with the correct auxiliary verbs.

Yes, I agree. It's probably better to do it properly or not at all. I wonder if @SaltfishAmi has any thoughts since I think you added vs-c handling in #450?

The vz class refers to zuru verbs, which are basically rendaku-ed suru verbs. Modern じる-verbs, such as 感じる, evolved from suru verbs (感する) to zuru verbs (感ずる), and eventually to ichidan verbs (感じる). Some forms of zuru verbs, like 感ぜられる, are still used in modern Japanese, often to evoke a classical style. Maybe, from a modern perspective, we can think of ずる-verbs as an alternative inflection pattern for じる-verbs.

Wow, thank you. I didn't know that background.

Should this be made more explicit?

No, that's perfect. Sorry, I only skimmed the tests but now I see the ずる tests etc.

[SaltfishAmi]

irregular / 不規則

simp. chinese: 不规则（same kanji but in simplified form）

I wonder if @SaltfishAmi has any thoughts since I think you added vs-c handling in #450?

I absolutely cannot recall what I did four years ago...

[enellis]

simp. chinese: 不规则（same kanji but in simplified form）

Very nice, thank you!

[birtles]

Thank you!