User:Zfe
Contents
What I am doing about Apertium
Tr-Az: A Savage Journey to the Heart of the Turanist Dream
After working a bit over en-tr, I decided to move to tr-az. Turkish and Azerbaijani are pretty similar languages and I'm confident to achieve great results within months! </ambition>
What we need to do for tr-az
1. Azerice needs a morph analyzer, we are working on it! It is called azmorph (Yes, pretty boring as a name, a new and exciting name coming soon!), it works pretty well for now, it can handle t_aor_present (which is the only present of azerice), vowel harmony, past
2. Dix: We need a decent bidix!
3. Some rules and exceptions: even though tr and az are really similar languages there will be differences to handle! We are gonna do it with CG and exceptions
TODOLIST
Trmorph
- Fix punctuation -- 3271696
- Proper names in caps
- Subcategorise proper names as far as possible
- Interrogative "mi" is broken
Azmorph
- Add consonant harmony |--| Azerbaijani, beside having double vowel harmony like turkish, has consonant harmony. Q/K (as well as their devoiced version ğ/y) change according to the precedent vowel.
- Remove 040-exception_ben.fst |--| Unlike turkish, azerbaijani doesn't have irregular dative for men and sen (personal pronouns).
- Remove 050-exception_su.fst |--| Su, the only irregular noun in turkish, is not irregular in azerbaijani.
- Disambiguation sucks big time: I don't know why it doesn't take içerler as <v><t_aor><3pp> but as a noun. Need to fix ASAP
- Fix punctuation
- Proper names in caps
- Subcategorise proper names as far as possible
- Interrogative mi