Zazaki and English/Ambiguity Resolution

Zazaki

Initial evaluation of ambiguity

• Basic corpus: 1.027
• Luke ch. 1–8: 1.023

There is a small amount of ambiguity in the transducer.

What forms are ambiguous?

Some words in Zazaki have multiple analyses:

• bi : conjunction/preposition "with" (bi<cnjcoo>) / past tense 3rd person masculine singular of "to be" (bîy<v><past><p3><m><sg>)
• ew : 3rd person singular pronoun "he/she" (ew<prn><pers><p3><sg><dir>) vs. 3rd person plural pronoun "they" (ew<prn><pers><p3><pl><dir>)
• Masculine nouns ending in -î : plural direct / singular oblique (e.g. mêrikî, silametî, qelebalixî, kesî)

Identified ambiguity

1. bi (preposition "with" vs. verb "was"):

In the following sentence, bi is the preposition "with" (before a noun):

^Şagirdan/*Şagirdan$ ^ra/ra<post>$ ^Yaqûb/*Yaqûb$ ^û/û<cnjcoo>$ ^Yûhenayî/*Yûhenayî$ ... ^ma/ma<prn><pers><p1><pl>$ ^bi/bi<cnjcoo>/bîy<v><past><p3><m><sg>$ ^yew/yew<n><m><sg><dir>$ ^emir/emir<n><m><sg><dir>$ ...

In the following sentence, bi is the verb "was" (before punctuation):

^Nameyê/*Nameyê$ ^ey/ey<adv>$ ^Zekerîya/*Zekerîya$ ^bi/bi<cnjcoo>/bîy<v><past><p3><m><sg>$^,/,<cm>$

Both sentences are from the Luke corpus.

Disambiguation rules:

• If bi is followed by a nominal (noun, determiner, adjective, numeral, or pronoun), select the conjunction/preposition reading.
• If bi is followed by sentence ending punctuation or a comma, select the verb reading.

2. ew (singular "he/she" vs. plural "they"):

In the following sentence, ew is singular (she), because the verb bena is feminine singular:

^ew/ew<prn><pers><p3><pl><dir> / ew<prn><pers><p3><sg><dir>$ ^zu/*zu$ ^way/*way$ ^cı/*cı$ ^bena/bîy<v><pres><p3><f><sg>$

Disambiguation rules:

• If ew is followed anywhere in the sentence by a singular verb, select the singular reading.
• If ew is followed by a plural verb form, select the plural reading.

Both sentences are from the basic corpus.

Final evaluation of ambiguity

• Basic corpus: 1.019 (down from 1.027)
• Luke ch. 1–8: 1.018 (down from 1.023)

GitHub repo

Find the transducer repo and language pair repo here.

Zazaki → English

Lexical selection

roj: "day" vs. "sun"

In Zazaki, roj (masculine noun) can mean both "day" and "sun". English distinguishes these as two separate words, but Zazaki uses one word for both.

Example sentences:

roj = "day":

Ê rojan de Sezar Agustosî emir kerd ke heme erdanê binê hakimîyetê Roma de nifûs bêro hûmaritiş.
"In those days Caesar Augustus ordered that a census be taken in all the lands under Roman rule."
(From the Luke corpus.)

roj = "day" (with "night"):

Roj û şewe roje girewtêne û Homayî rê duayî kerdêne û îbadet kerdêne.
"Day and night she was fasting and praying and worshipping God."
(From the Luke corpus.)

roj = "sun":

Roj gırd bi.
"The sun was big."
(Constructed example.)

Generalisation: The default translation is "day", as this is the most common usage in the corpus. When roj appears after a numeral or alongside time-related words like şewe ("night"), the "day" reading is reinforced. The "sun" reading would apply in contexts about sky, light, or weather, though these are rare in the current corpus.

Rules implemented:

• Default rule (low weight): roj → day
• After a numeral: roj → day (high weight)
• Before şewe ("night"): roj → day (high weight)

Bilingual dictionary: Two entries for roj in apertium-zza-eng.zza-eng.dix, one mapping to "day" and one to "sun".

Testing:

$ echo "roj" | apertium -d . zza-eng-biltrans
day/sun  (both options shown)
$ echo "roj" | apertium -d . zza-eng
day  (lexical selection chooses "day")

Find the transducer repo and language pair repo here.