mjohnson/espeak-ng
1
0
Fork 0
mirror of https://github.com/rhasspy/espeak-ng.git synced 2026-10-31 13:38:47 +00:00
Code Issues Packages Projects Releases Wiki Activity

Move handling of SetLetterVowel() to language files

Browse Source 
Contributes to:
Issue #218, https://github.com/espeak-ng/espeak-ng/issues/218

Changes:
language files have a new keyword letterVowel. It can be a latin alphabet from a-z or a hex value (used by bulgarian).
Errors in parsing the values are only reported, nothing is done about them.

About testing:
I haven't noticed any differences in the output audio with letterVowel set or unset in any tested language. The code seems to work and the correct bits seem to be set, but I don't know how to confirm from the final audio.

TODO:
1. Write better documentation in docs/voices.md
2. The code uses new_translator. Should it use translator instead?
This commit is contained in:
Juho Hiltunen
2017-12-10 21:04:36 +02:00
parent 4f56345fb5
commit 6042eb8d1e
27 changed files with 68 additions and 31 deletions
Download Patch File Download Diff File Expand all files Collapse all files
docs/voices.md
+7 -1
View File
@@ -30,7 +30,7 @@
- [stressAmp](#stressamp)
- [intonation](#intonation)
- [dictmin](#dictmin)
- [letterVowel](#letterVowel)
----------
A Voice file specifies a language (and possibly a language variant or
@@ -425,3 +425,9 @@ Used for some languages to detect if additional language data is
installed. If the size of the compiled dictionary data for the language
(the file `espeak-ng-data/*_dict`) is less than this size then a
warning is given.
### letterVowel
letterVowel <letter>
Used for some languages to handle a certain letter as a vowel instead of consonant.
espeak-ng-data/lang/art/jbo
+1
View File
@@ -2,3 +2,4 @@ name Lojban
language jbo
speed 80 // speed adjustment, percentage
letterVowel y
espeak-ng-data/lang/cel/cy
+2
View File
@@ -2,3 +2,5 @@ name Welsh
language cy
intonation 4
letterVowel w
letterVowel y
espeak-ng-data/lang/gmq/da
+1
View File
@@ -2,3 +2,4 @@ name Danish
language da
tunes s2 c2 q2 e2
letterVowel y
espeak-ng-data/lang/gmq/is
+1
View File
@@ -1,2 +1,3 @@
name Icelandic
language is
letterVowel y
espeak-ng-data/lang/gmq/nb
+1
View File
@@ -5,3 +5,4 @@ phonemes no
dictionary no
intonation 4
letterVowel y
espeak-ng-data/lang/gmq/sv
+1
View File
@@ -1,2 +1,3 @@
name Swedish
language sv
letterVowel y
espeak-ng-data/lang/gmw/af
+1
View File
@@ -6,3 +6,4 @@ status mature
roughness 0
pitch 63 120
letterVowel y
espeak-ng-data/lang/gmw/de
+2
View File
@@ -1,2 +1,4 @@
name German
language de
letterVowel y
espeak-ng-data/lang/gmw/nl
+1
View File
@@ -1,2 +1,3 @@
name Dutch
language nl
letterVowel y
espeak-ng-data/lang/ine/sq
+1
View File
@@ -1,5 +1,6 @@
name Albanian
language sq
letterVowel y
// add this line to remove 'ë' at the end of words
// replace 00 @/ NULL
espeak-ng-data/lang/roa/fr
+1
View File
@@ -4,3 +4,4 @@ language fr
dictrules 1
tunes s3 c3 q3 e3
letterVowel y
espeak-ng-data/lang/roa/fr-BE
+1 -1
View File
@@ -4,5 +4,5 @@ language fr 8
dictrules 2
tunes s3 c3 q3 e3
letterVowel y
espeak-ng-data/lang/roa/fr-CH
+1
View File
@@ -4,3 +4,4 @@ language fr 8
dictrules 3
tunes s3 c3 q3 e3
letterVowel y
espeak-ng-data/lang/roa/it
+1
View File
@@ -5,3 +5,4 @@ maintainer Christian Leo M <llajta2012@gmail.com>
status mature
tunes s4 c4 q4 e4
letterVowel y
espeak-ng-data/lang/roa/pt
+1
View File
@@ -5,3 +5,4 @@ phonemes pt-pt
dictrules 1
intonation 2
letterVowel y
espeak-ng-data/lang/urj/fi
+1
View File
@@ -1,2 +1,3 @@
name Finnish
language fi
letterVowel y
espeak-ng-data/lang/urj/hu
+1 -1
View File
@@ -2,5 +2,5 @@ name Hungarian
language hu
option bracket 0 0
pitch 81 117
letterVowel y
espeak-ng-data/lang/zls/bg
+2
View File
@@ -3,3 +3,5 @@ language bg
stressAmp 13 12 17 17 20 22 22 21
stressLength 180 170 200 200 200 200 210 220
letterVowel 0x2a
espeak-ng-data/lang/zls/bs
+2
View File
@@ -12,3 +12,5 @@ formant 5 97 102 100
stressAdd 10 10 0 0 0 0 -30 -30
dictrules 3 4
letterVowel y
letterVowel r
espeak-ng-data/lang/zls/hr
+2
View File
@@ -13,3 +13,5 @@ formant 5 97 102 100
stressAdd 10 10 0 0 0 0 -30 -30
dictrules 1
letterVowel y
letterVowel r
espeak-ng-data/lang/zls/sr
+3
View File
@@ -11,3 +11,6 @@ formant 5 97 102 100
stressAdd 10 10 0 0 0 0 -30 -30
dictrules 2 4
letterVowel y
letterVowel r
espeak-ng-data/lang/zlw/cs
+2
View File
@@ -1,2 +1,4 @@
name Czech
language cs
letterVowel y
letterVowel r
espeak-ng-data/lang/zlw/pl
+1
View File
@@ -2,3 +2,4 @@ name Polish
language pl
intonation 2
letterVowel y
espeak-ng-data/lang/zlw/sk
+2
View File
@@ -1,2 +1,4 @@
name Slovak
language sk
letterVowel y
letterVowel r
src/libespeak-ng/tr_languages.c
-27
View File
@@ -121,11 +121,6 @@ ALPHABET *AlphabetFromChar(int c)
static void Translator_Russian(Translator *tr);
static void SetLetterVowel(Translator *tr, int c)
{
tr->letter_bits[c] = (tr->letter_bits[c] & 0x40) | 0x81; // keep value for group 6 (front vowels e,i,y)
}
static void ResetLetterBits(Translator *tr, int groups)
{
// Clear all the specified groups
@@ -483,7 +478,6 @@ Translator *SelectTranslator(const char *name)
tr->langopts.vowel_pause = 0x30;
tr->langopts.param[LOPT_DIERESES] = 1;
tr->langopts.param[LOPT_PREFIXES] = 1;
SetLetterVowel(tr, 'y'); // add 'y' to vowels
tr->langopts.numbers = NUM_SWAP_TENS | NUM_HUNDRED_AND | NUM_SINGLE_AND | NUM_ROMAN | NUM_1900;
tr->langopts.accents = 1;
@@ -512,7 +506,6 @@ Translator *SelectTranslator(const char *name)
case L('b', 'g'): // Bulgarian
{
SetCyrillicLetters(tr);
SetLetterVowel(tr, 0x2a);
tr->encoding = ESPEAKNG_ENCODING_ISO_8859_5;
tr->langopts.param[LOPT_UNPRONOUNCABLE] = 0x432; // [v] don't count this character at start of word
tr->langopts.param[LOPT_REGRESSIVE_VOICING] = 0x107; // devoice at end of word, and change voicing to match a following consonant (except v)
@@ -581,8 +574,6 @@ Translator *SelectTranslator(const char *name)
tr->langopts.numbers = NUM_OMIT_1_HUNDRED;
SetLetterVowel(tr, 'w'); // add letter to vowels and remove from consonants
SetLetterVowel(tr, 'y');
}
break;
case L('d', 'a'): // Danish
@@ -592,7 +583,6 @@ Translator *SelectTranslator(const char *name)
tr->langopts.stress_rule = STRESSPOSN_1L;
tr->langopts.param[LOPT_PREFIXES] = 1;
SetLetterVowel(tr, 'y');
tr->langopts.numbers = NUM_DECIMAL_COMMA | NUM_SWAP_TENS | NUM_HUNDRED_AND | NUM_OMIT_1_HUNDRED | NUM_ORDINAL_DOT | NUM_1900 | NUM_ROMAN | NUM_ROMAN_CAPITALS | NUM_ROMAN_ORDINAL;
}
break;
@@ -609,7 +599,6 @@ Translator *SelectTranslator(const char *name)
tr->langopts.param[LOPT_LONG_VOWEL_THRESHOLD] = 175/2;
tr->langopts.numbers = NUM_DECIMAL_COMMA | NUM_SWAP_TENS | NUM_ALLOW_SPACE | NUM_ORDINAL_DOT | NUM_ROMAN;
SetLetterVowel(tr, 'y');
tr->langopts.param[LOPT_UNPRONOUNCABLE] = 2; // use de_rules for unpronouncable rules
}
break;
@@ -801,7 +790,6 @@ Translator *SelectTranslator(const char *name)
tr->langopts.long_stop = 130;
tr->langopts.numbers = NUM_DECIMAL_COMMA + NUM_ALLOW_SPACE;
SetLetterVowel(tr, 'y');
tr->langopts.spelling_stress = 1;
tr->langopts.intonation_group = 3; // less intonation, don't raise pitch at comma
}
@@ -816,7 +804,6 @@ Translator *SelectTranslator(const char *name)
tr->langopts.accents = 2; // Say "Capital" after the letter.
tr->langopts.numbers = NUM_SINGLE_STRESS | NUM_DECIMAL_COMMA | NUM_ALLOW_SPACE | NUM_OMIT_1_HUNDRED | NUM_NOPAUSE | NUM_ROMAN | NUM_ROMAN_CAPITALS | NUM_ROMAN_AFTER | NUM_VIGESIMAL | NUM_DFRACTION_4;
SetLetterVowel(tr, 'y');
}
break;
case L('g', 'a'): // irish
@@ -900,8 +887,6 @@ Translator *SelectTranslator(const char *name)
tr->langopts.replace_chars = replace_cyrillic_latin;
tr->langopts.our_alphabet = OFFSET_CYRILLIC; // don't say "cyrillic" before letter names
SetLetterVowel(tr, 'y');
SetLetterVowel(tr, 'r');
}
break;
case L('h', 't'): // Haitian Creole
@@ -929,7 +914,6 @@ Translator *SelectTranslator(const char *name)
tr->langopts.decimal_sep = ',';
tr->langopts.max_roman = 899;
tr->langopts.min_roman = 1;
SetLetterVowel(tr, 'y');
tr->langopts.spelling_stress = 1;
SetLengthMods(tr, 3); // all equal
}
@@ -986,7 +970,6 @@ Translator *SelectTranslator(const char *name)
SetLetterBits(tr, 4, "kpst"); // Letter group F
SetLetterBits(tr, 3, "jvr"); // Letter group H
tr->letter_groups[1] = is_lettergroup_B;
SetLetterVowel(tr, 'y');
tr->langopts.numbers = NUM_DECIMAL_COMMA | NUM_SINGLE_AND | NUM_HUNDRED_AND | NUM_AND_UNITS | NUM_1900;
tr->langopts.numbers2 = 0x2;
}
@@ -1011,7 +994,6 @@ Translator *SelectTranslator(const char *name)
tr->langopts.numbers2 = NUM2_NO_TEEN_ORDINALS;
tr->langopts.roman_suffix = utf8_ordinal;
tr->langopts.accents = 2; // Say "Capital" after the letter.
SetLetterVowel(tr, 'y');
}
break;
case L_jbo: // Lojban
@@ -1024,7 +1006,6 @@ Translator *SelectTranslator(const char *name)
tr->langopts.vowel_pause = 0x20c; // pause before a word which starts with a vowel, or after a word which ends in a consonant
tr->punct_within_word = jbo_punct_within_word;
tr->langopts.param[LOPT_CAPS_IN_WORD] = 2; // capitals indicate stressed syllables
SetLetterVowel(tr, 'y');
tr->langopts.max_lengthmod = 368;
}
break;
@@ -1195,7 +1176,6 @@ Translator *SelectTranslator(const char *name)
tr->langopts.param[LOPT_DIERESES] = 1;
tr->langopts.param[LOPT_PREFIXES] = 1;
tr->langopts.param[LOPT_REGRESSIVE_VOICING] = 0x100; // devoice at end of word
SetLetterVowel(tr, 'y');
tr->langopts.numbers = NUM_DECIMAL_COMMA | NUM_SWAP_TENS | NUM_OMIT_1_HUNDRED | NUM_OMIT_1_THOUSAND | NUM_ALLOW_SPACE | NUM_1900 | NUM_ORDINAL_DOT;
tr->langopts.ordinal_indicator = "e";
@@ -1209,7 +1189,6 @@ Translator *SelectTranslator(const char *name)
SetupTranslator(tr, stress_lengths_no, NULL);
tr->langopts.stress_rule = STRESSPOSN_1L;
SetLetterVowel(tr, 'y');
tr->langopts.numbers = NUM_DECIMAL_COMMA | NUM_HUNDRED_AND | NUM_ALLOW_SPACE | NUM_1900 | NUM_ORDINAL_DOT;
}
break;
@@ -1240,7 +1219,6 @@ Translator *SelectTranslator(const char *name)
tr->langopts.numbers = NUM_DECIMAL_COMMA | NUM_ALLOW_SPACE | NUM_DFRACTION_2;
tr->langopts.numbers2 = NUM2_THOUSANDS_VAR3;
tr->langopts.param[LOPT_COMBINE_WORDS] = 4 + 0x100; // combine 'nie' (marked with $alt2) with some 1-syllable (and 2-syllable) words (marked with $alt)
SetLetterVowel(tr, 'y');
}
break;
case L('p', 't'): // Portuguese
@@ -1256,7 +1234,6 @@ Translator *SelectTranslator(const char *name)
tr->langopts.numbers = NUM_DECIMAL_COMMA | NUM_DFRACTION_2 | NUM_HUNDRED_AND | NUM_AND_UNITS | NUM_ROMAN_CAPITALS;
tr->langopts.numbers2 = NUM2_MULTIPLE_ORDINAL | NUM2_NO_TEEN_ORDINALS | NUM2_ORDINAL_NO_AND;
tr->langopts.max_roman = 5000;
SetLetterVowel(tr, 'y');
ResetLetterBits(tr, 0x2);
SetLetterBits(tr, 1, "bcdfgjkmnpqstvxz"); // B hard consonants, excluding h,l,r,w,y
tr->langopts.param[LOPT_ALT] = 2; // call ApplySpecialAttributes2() if a word has $alt or $alt2
@@ -1314,8 +1291,6 @@ Translator *SelectTranslator(const char *name)
if (name2 == L('c', 's'))
tr->langopts.numbers2 = 0x108; // variant numbers before milliards
SetLetterVowel(tr, 'y');
SetLetterVowel(tr, 'r');
ResetLetterBits(tr, 0x20);
SetLetterBits(tr, 5, sk_voiced);
}
@@ -1368,7 +1343,6 @@ Translator *SelectTranslator(const char *name)
tr->langopts.stress_rule = STRESSPOSN_1R;
tr->langopts.stress_flags = S_FINAL_DIM_ONLY | S_FINAL_NO_2 | S_FINAL_VOWEL_UNSTRESSED;
SetLetterVowel(tr, 'y');
tr->langopts.numbers = NUM_DECIMAL_COMMA | NUM_HUNDRED_AND | NUM_AND_UNITS | NUM_DFRACTION_4;
tr->langopts.accents = 2; // "capital" after letter name
}
@@ -1380,7 +1354,6 @@ Translator *SelectTranslator(const char *name)
SetupTranslator(tr, stress_lengths_sv, stress_amps_sv);
tr->langopts.stress_rule = STRESSPOSN_1L;
SetLetterVowel(tr, 'y');
tr->langopts.numbers = NUM_SINGLE_STRESS | NUM_DECIMAL_COMMA | NUM_ALLOW_SPACE | NUM_1900;
tr->langopts.accents = 1;
}
src/libespeak-ng/voices.c
+28 -1
View File
@@ -109,7 +109,11 @@ enum {
// these need a phoneme table to have been specified
V_REPLACE,
V_CONSONANTS
V_CONSONANTS,
// these are alpha features that need to be tested and categorized
V_LETTER_VOWEL
};
static MNEM_TAB options_tab[] = {
@@ -168,6 +172,8 @@ static MNEM_TAB keyword_tab[] = {
{ "l_length_mods", 0x100+LOPT_LENGTH_MODS },
{ "apostrophe", 0x100+LOPT_APOSTROPHE },
// these are alpha features that need to be tested and categorized
{ "letterVowel", V_LETTER_VOWEL },
{ NULL, 0 }
};
@@ -858,6 +864,27 @@ voice_t *LoadVoice(const char *vname, int control)
case V_MAINTAINER:
case V_STATUS:
break;
case V_LETTER_VOWEL: {
char str[5] = "";
char c = '0';
char *endptr = NULL;
sscanf(p, "%s", str);
// assume a hex value if string starts with "0x"
if (str[0] == '0' && str[1] == 'x') {
c = strtoul(str, &endptr, 16);
if (errno == ERANGE)
fprintf(stderr, "letterVowel out of range.\n");
}
else {// otherwise, assume a single letter
c = str[0];
if (c < 97 || c > 122) // valid values are a-z, ascii 97-122
fprintf(stderr, "letterVowel out of range.\n");
}
new_translator->letter_bits[c] = (new_translator->letter_bits[c] & 0x40) | 0x81; // keep value for group 6 (front vowels e,i,y)
break;
}
default:
if ((key & 0xff00) == 0x100) {
if (langopts)