Felicia ªerban & al. * Database of the Romanian Language Phonetics and Phonology

One has less problems with the consonants, but still we have to mention some of them.

• The letters c and g are used in three ways, just like in Italian: (1) followed by certain vowels such as [a, o, u], they represent the velar occlusive phonemes [k] and [g] (caz 'case', colorat 'coloured', curte 'court', gol 'empty', gust 'taste'); (2) followed by /i/ or /e/, the letters c and g represent palato-alveolar affricates (gen 'gender', gin); in such a situation, the letters c and g followed by /i/ or /e/ may sometimes represent two phonemes and sometimes only one phoneme: cinci 'five' = CVCC$, giuvaergiu 'jeweller' = C$V/CV/VC/CWS; (3) followed by h and by the vowels /e/ or /i/, they compose the monophonematic groups of consonants ch = [k], gh = [g] (a chema 'to call', ghitarist 'guitar player') [3,4].
  
  
• The letter x, used only in neologisms, is biphonematic, read like in French: (1) sometimes [k+s] (boxer 'pugilist'); (2) other times [g+z] (examen 'examination'). In this case, the problems refer only to pronunciation, without further consequences on syllabication; during flexion, the words ending in [k+s] present a consonantic alternation, the two phonemes being written by means of different letters (complex CVC/CCWC, masculine plural form CVC/CCWCCP). We have to point out that the same phonemes /k+s/ can sometimes be represented by two letters (cocs 'coke', sconcs 'skunk').

2.1.2. Vowels

Vowels and semivowels cause the greatest problems in the automatic mapping of letters onto phonemes, with consequences on syllabication. For some of these, the reference literature [3] offers information which must be checked (especially in the case of neologisms) and completed. The cause is that not the whole database stock was examined from the very beginning.

• Romanian language has seven vowels and all can form syllables on their own or together with one (or more) consonants. From these, [a] and other two vowels from the central series always form a syllable of their own.
  
  
• In turn, /e/, /i/, /o/ and /u/ are sometimes syllabic and other times, even though they are represented by the same graphic signs, semivocalic (S), forming diphthongs together with a vowel from the previous series or combining among themselves [5]: seara 'evening' CSW/CV, ianuarie 'January' SV/CV/W/CV/V, soare 'sun' CSW/CV, cuantum 'quantum' CSWC/CVC, mai 'May' CVS, sau 'or' CVS, ardei 'red pepper' VC/CWS, iepure 'hare' SW/CV/CV, iod 'iodine' SVC, apoi 'then' V/CWS, cadou 'gift' CV/CWS, mereu 'continually' CV/CWS, detaliu 'detail' CV/CW/CSV, auriu 'golden' V/V/CWS. Two semivowels may combine with a vowel, forming triphthongs: turcoaice 'Turkish women' CVC/CSWS/CV, iau 'I take' SWS, iei 'you take' SWS.
  
  
• Besides a vocalic /i/ and a semivocalic one, the letter i may represent in Romanian a pseudovowel (P), a soundless /i/ [6], which marks the palatalization of the preceding consonant [7], in fact a short /i/ [8], usually at the end of the word, in plural forms or in some flexional forms of the verb: corbi CVCCP < corb 'raven', lumi CVCP < lume 'world', luni 'Monday' CVCP, vulpi CVCCP < vulpe 'fox', mori 'you die' CVCP < muri 'to die' CV/VW, and so on. As one can notice, not any final postconsonantic /i/ is devocalised. Sometimes, a final /i/ may form a syllable even if it is not stressed (e.g. after two consonants, when the last one is a liquid [l] or [r]: codri CW/CCV < codru 'forest', afli 'you find' W/CCV < afla 'to find' V/CCW). There are other situations, too, such as centi 'centimetre' CWC/CV, confeti 'confetti' CVC/CW/CV, tanti 'aunt' CWC/CV, and so on. It must be added that a devocalised /i/ may not appear only at the end of the word but also within it, in compound words: oricum 'anyway' VCP/CWC.
  
  
• Finally, besides the above-mentioned values, the letters e and i after c and g do not have a phoneme value (noted by $); they show that the consonants are not the occlusives [k] and [g], but palato-alveolar affricates: cioban 'shepherd' C$V/CWC, geamgiu 'glazier' C$VC/CWS, giulgiu 'shroud' C$WC/C$V. It is difficult to establish rules for mapping the letters onto phonemes for the purpose of automatic syllabication: geometrie 'geometry' CV/V/CV/CCW/V, but georgian C$VC/CV/WC.

2.2. Automatic syllabication

In order to syllabicate the words in Romanian, the reference normative studies accept phonetic syllabication but also recommend the usage of morphological syllabication (which takes into account the internal structure of the word).

The syllabication rules were formalised for the computer requests. Research is facilitated by the 'filter' included in the soft, which allows the user to select certain groups of letters (coded or in clear) and to extract all the words of the database which contain that sequence. It is possible to choose the words which contain it at the beginning, at the end, or in any other position.

189