* Dict
- Dict only contains nouns and adjectives using the full-form
paradigms^1 (for stability and because generating the minimal ones
would take longer)
- verbs can be found in Irreg, using bescherelle paradigms -- the
full-form ones would make the file too big!
- adverbs are just {s : Str}, so if you know their names you can make
them yourself, no need to include them (but I can if it's best)
^1: adjectives actually use the almost full-form paradigm, but they
should be correct except for a couple of cases
* Irreg
- renamed IrregBesch to Irreg
- added more verbs (there are still quite a few missing)
(Ara) Add new mkV2V constructors + better comments for degrA
(Ara) Add new paradigm for assimilated+defective FormI verbs
(Ara) misc. small fixes/additions
- make it about guessing feminine form from the lemma (masculine form)
- this way one can reuse the noun paradigm in the adjective paradigms,
simplifying it and improving it at the same time
- add cases for 'mente'
obs: works but doesn't compile?
ImpersCl (VP -> Cl) used to check whether the VP is predicative, and
in the positive case, force the dummy subject to not be a pronoun.
This has been wrong since the time I changed prodrop rules in PredVP.
The error that followed manifested when the Cl was put in subordinate
position: the subject pronoun did not attach to the conjunction.
An example follows.
Lang: SubjS that_Subj (UseCl (TTAnt TPres ASimul) PPos (ImpersCl (UseComp (CompNP (DetCN (DetQuant IndefArt NumSg) (UseN girl_N))))))
LangEng: that it is a girl
Old LangAra: أنَّ ها بِنتٌ (wrong)
New LangAra: أنَّ &+ ها بِنتٌ (correct)
This makes a difference in word order; default is VSO, but if the
subject NP is complex, e.g. coming from RelNP or SentCN,
the more natural word order is SVO.
Example that triggered the change comes from an application grammar:
Gold standard by informant: مَا تَوَصْلّنَا إلَيهِ يَبْدُو مَنْطِقِيَّاً
Grammar before change: يَشْعُرُ مَاً نَتَوَصَّلُ إِلَىهُ مَنْطِقِيّاً
Grammar after change: مَاً نَتَوَصَّلُ إِلَىهُ يَشْعُرُ مَنْطِقِيّاً