1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
|
LL AAA 111
LL AA AA 1111
LL AA AA 11 11
LL AA AA 11
LL AA AA 11
LL AAAAAAAAAAA 11
LL AA AA 11
LL AA AA 11
LLLLLLLLL AA AA 111111 June 05
--------------------------------------------------------------
LA1 - Classical Latin male voice for the MBROLA speech synthesizer
Created by Olivier Bianchi
Laboratory for the Computational Analysis of Speech
University of Lausanne
--------------------------------------------------------------
--------------------------------------------------------------
Table of Contents
--------------------------------------------------------------
1.0 Brief description of the LA1 diphone database
2.0 Files
3.0 Installation and tests
4.0 Announcement
5.0 Acknowledgements
--------------------------------------------------------------
1.0 Brief description of the LA1 diphone database
--------------------------------------------------------------
LA1 is a diphone database for classical Latin, consisting of
1275 diphones, male voice, and was created in the context of
the author's doctoral project.
LA1 provide diphone for classical Latin pronunciation only
(i.e. 1st century BC).
LA1 uses the SAMPA phonetic alphabet as defined under
http://www.phon.ucl.ac.uk/home/sampa/home.htm. In addition to
the Latin phonemes, some Greek phonemes are included in order
to allow synthesis of Greek words transcripted in Latin caracter.
Long and short vowels are considered different, according to the
newest phonological theories. However, not dinstinction was made
between stressed and unstressed vowels or diphtongs.
Double (geminate) consonants are considerated as the correspondingly
lengthened sound. Double consonants are included in diphone sets
to improve the acoustical database quality.
For further information about phoneme list, see:
Bianchi, O., Orator Romanus. Une synthèse de la parole pour le
latin classique, Université de Lausanne, 2005. (IN PROGRESS)
There are six 'simple' plosives, p b t d k g, and six double
plosives, p: b: t: d: k: g:, and three simple aspirated voiceless
plosives, p_h t_h k_h, and three double aspirated plosives,
p_h: t_h: k_h:, for Greek words transcription:
SAMPA Graph Like in
p p pater, septem
b b bene
t t et, terror
d d duco, fides
k c/k actus, cloaca, kalendae
g g flagrum, reges
p: pp cuppa
b: bb abbreuio
t: tt mitto
d: dd reddere
k: cc assiccare
g: gg aggero
p_h ph philosophia
t_h th Cethegus
k_h ch chorus
p_h: ph Sapphus
t_h: th leget Thyni
k_h: ch haec chrysoprys
There are three simple fricatives, f s h, and two double frivatives,
f: s:, and one simple fricative, z, and one double fricative, z:,
for Greek words transcription:
f f facile, infra
s s soror, filius
h h- hoc
z z- zema
f: ff differo
s: s/ss causa/fissus
z: -z- zizania
There are three simple nasals, m n N, and two double nasals, m: n:,
two simple liquids, l r, and two double liquids, l: r:, and two
simple semivowels, w j, and one double semivowel, j: :
m m amare, manere
n n nox, lumen
N g/n regnum, relinquo
m: mm accommodus
n: nn annus
l l labor, uel
r r gero, ratio
l: ll tabella
r: rr terra
w u/v uoco/voco
j i/j iam/jam
j: i/j aio/Ajax
There are five short vowels, a E I O U, and one short vowel for
Greek words transcription, y. Except the central vowel a and the Greek
vowel y, the short vowel appear to have been appreciably more open than the
long:
a a caput, amans
E e pulcher, epulae
I i eximo, ille
O o censor, optimus
U u amicus, gutta
y y cytisus
There is also an 'intermediate vowel' between I et U, 1, that appear only
in particular environments:
1 i/u optimus/optumus
There are five long vowels, a: e: i: o: u:, and one long vowel for
Greek words transcription, y:, and three diphtongs, aE aU OE :
a: a latrina, ingratia
e: e institutiones, uolebam
i: i boni, audire
o: o impono
u: u quamdiu, sternutare
y: y Hylaeus, gyrus
aE ae rosae, Caesar
aU au adauctare, faustus
OE oe foedus
--------------------------------------------------------------
2.0 Files
--------------------------------------------------------------
la1.txt : this file
la1 : the diphone database
test_1.pho : Example : "Multae res ad hoc consilium Gallos
hortabantur."
test_2.pho : Another example : "Nullum aditum esse ad eos
mercatoribus."
--------------------------------------------------------------
3.0 Installation and Tests
--------------------------------------------------------------
[part mostly cut&paste from MBROLA help files]
If you have not copied the MBROLA software yet, please consult
the MBROLA project homepage and get it at
http://tcts.fpms.ac.be/synthesis/mbrola.html
Copy la1.zip into the mbrola directory and unzip it :
unzip la1.zip (or pkunzip on PC/DOS)
(don't forget to delete the .zip file when this is done)
Try
mbrola la1 test/test_1.pho test_1.wav
to create a sound file for a phrase. In this example the
audio file follows the RIFF Wav format. But depending on the
extension test.au, test.aif, or test.raw you can obtain
other file formats. Listen to it with your favorite sound editor,
and try the other command files (*.pho) to have a better idea of
the quality of speech you can synthesize with the MBROLA technique.
On Unix systems you can pipe the audio ouput to the sound player as
on a HP : mbrola LA1 test/test_1.pho - | splayer -srate 16000 -l16
--------------------------------------------------------------
4.0 Announcement
--------------------------------------------------------------
This work is only a part of the whole doctoral project, whose goal
is to provide high quality classical Latin speech synthesis, called
Orator Romanus (Laiptts_L). This system, which runs on top of
MBROLA and the LA1 database, included rules for duration
and pitch control.
For more information contact the author at olivier.bianchi@unil.ch
--------------------------------------------------------------
5.0 Acknowledgments
--------------------------------------------------------------
This database was recorded by Tomas Dubeda, a Czech researcher.
Thank you very much Tomas!
We would like to thank Baris Bozkurt for his efforts in converting
our diphone database to the MBROLA format.
--------------------------------------------------------------
June 9, 2005
Olivier Bianchi
For more information contact: olivier.bianchi@unil.ch
--------------------------------------------------------------
|
