-*- mode: outline -*-
The corpus for Project Santiago, Croatian in two alphabets

* Introduction
The following is a design analysis of the corpus made for Project
Santiago to be used in Croatian by speakers reading one text in two
alphabets, Latin and Cyrillic. It is interesting to note that, while
reference will be made to a single language, within that referenced
language and enormous set of natural variants occurs.
**Dialect
While there is no single dialect of the spoken language that could be
safely considered the most prevalent, the ijekavski dialect has been
chosen as representative of a large number of Croatian speakers with
significant overlap into Bosnian and Serbian dialects. The term -ije
derives from the manifestation of the -e vowel (formerly jat, in Old
Church Slavic). The ekavski dialect is rare within Croatia, but another
dialect, the ikavski, prevails, especially along the Dalmatian coast.
Speakers from these other dialects will likewise replace the sto in the
stokavian dialect with sta; ca or kaj are also possible especially
within the ijekavski variant.

. Certain changes in the dialect revert to the consonants, causing a
palatalization. Pesma in the ekavski dialect would be pjesma in the
-ijekavski variant. However, except for nj and lj no palatalized
consonant is phonemically recognized. Other palatalizations such as in
pesma/pjesma have been avoided for the sake of dialectal unity insofar
as that is possible.

Tonal registers in the accented vowel do not technically attend the
Croatian ijekavski dialect. Vocalic length is not considered phonemic
within the same speech group. However, many ijekavski speakers do use
tones and length within their vowels. The following corpus is too short
to give proper representation of all 36 accented and long combinations
in conjunction with all 24 consonants, a total of 864 CV possibilities,
exclusive of word initial vowels.

Since Croatian has retained the open syllable principle of early Slavic,
it was possible to use open syllables for the accented vowel occurrence,
which was judged optimal for modelling concerns. Resonants including
liquids, were also avoided when possible at syllable boundary to avoid
coarticulation. 

The Serbian graphic system has a one-to-one correspondence to the
Croatian or latin aplphabet. The digraphs Dzh , lj, and nj are all
single elements in the cyrillic. The latin script can be exchanged for
cyrillic without great modification.

*Croatian phones

** 1.0 Vowels

*** 1.1 There are six base vowels, /a/, /e/, /i/, /o/, /r/, /u/ . The /e/ in
some dialects alternates with /i/ (ikavski dialect) or /ije/ (ijekavski
dialect). Some /e/ sounds remain intact within those dialects, depending
on the history of the word.

*** 1.2 This corpus represents the ijekavski dialect. Technically, /ije/ is
one long vowel. However, this corpus will treat the two syllables as two
vowels, using the standard assignation of rising or falling tones across
both vowels. Falling tone is marked on /i/, rising tone is marked on
/e/. Both these syllables are invariably short.

*** 1.3 All vowels retain the phonetic quality typical of accented vowels.
There are no diphthongs. In the enclitics,  je, se, the vowels are so
short quantitatively (as short as 50 ms.)  that it is often difficult to
assign qualitative measures to them at all; they sometimes barely exceed
 VOT [voice onset timing], for example). However, purists of the
language would not believe there is qualitative deterioration. In
careful speech or emphatic situations, hypercorrection or extenuation
occurs in these words, drawing them out to 80-90 ms, the length more
typical of an ordinary syllable with a short-unstressed vowel.

*** 1.4 All vowels have six possible realizations, two without stress and
two each of two different tonal stresses:

	

****        Unstressed

   v  - short 

   V - long 

**** 	        Stressed

***** 	  Falling         Rising

****** Short      v  V             v  V

****** Long      v   V            v  V

** 2.0 Consonants

*** 2.1 There are six pairs of consonants by voice: 
  /ca/ /dj/  
/ch/ /dzh/  
/f/  /v/  
/k/  /g/  
/p/  /b/  
/s/  /z/ 
/sh/ /zh/
  /t/   /d/  
*** 2.2 There are seven unpaired resonants, including two semi-vowels*:

/j/ 
/l/*  
/nj/ 
/m/ 
/n/ 
/nj/ 
/r/*    

*** 2.3 There are two unpaired, unvoiced consonants:

 /c/ 
/h/

** 3.0 Spelling conventions and phonetics

Croatian is spelled phonetically, despite the loss of morphemic
distinctions:

	On je tezhak.

	Ona je teshka.

*** 3.1 Assimilation across word boundaries can still take place when words
occur together in speech.