Numbers: Support "myriad" (10^4) based number systems (lang=ko).

Lang=pt: fix ordinal numbers 11-19.
Lang=en-us: minor vowel changes.


git-svn-id: https://espeak.svn.sourceforge.net/svnroot/espeak/trunk@291 d46cf337-b52f-0410-862d-fd96e6ae7743

dictsource/dict_phonemes

 Z    Z;   
 
 
-Dictionary ko_dict  2012-09-14
+Dictionary ko_dict  2012-09-15
 
 @    a    e    E    i    o    u    u-   
 

dictsource/ko_list

-$textmode
+// eSpeak Korean rules. Changed pronunciation, as numbers were not announced.
 
-// Number forms:
-_0      ᅧᆼ
-_1      ᅵᆯ
-_2      ᅵ
-_3      삼
-_4      사
-_5      ᅩ
-_6      ᅲᆨ
-_7      칠
-_8      팔
-_9      구
-_1X     십
-_2X     ᅵ십
-_3X     삼십
-_4X     사십
-_5X     ᅩ십
-_6X     ᅲᆨ십
-_7X     칠십
-_8X     팔십
-_9X     구십
-_0C     백
-_0M1    천
-_1M1    천
 
-$phonememode
-_0M2	m'an          // 10^4
-_1M2	m'an
-_0M3	tSh-@n||m'an  // 10^7
-_1M3	tSh-@n||m'an
-_0M4	'@q           // 10^8
-_1M4	'@q
-_0M5	tSh-@n||'@q   // 10^11
-_1M5	tSh-@n||'@q
-_0M6	tS;'o         // 10^12
-_1M6	tS;'o
-_0M7	tSh-@n||tS;'o
-_1M7	tSh-@n||tS;'o
-_0M8	gj'@nN
-_1M8	gj'@nN
-$textmode
+// Number forms:
+_0	j'@N_!
+_1	'iR_!
+_2	'i_!
+_3	s'am_!
+_4	s'a_!
+_5	'o_!
+_6	j'uq_!
+_7	tS'h'iR_!
+_8	ph'aR_!
+_9	g'u_!
+_1X	s'ip_!
+_2X	'i_!||s'ip_!
+_3X	s'am_!||s'ip_!
+_4X	s'a_!||s'ip_!
+_5X	'o_!||s'ip_!
+_6X	j'uq_!||s'ip_!
+_7X	tS'h'iR_!s'ip_!
+_8X	ph'aR_!||s'ip_!
+_9X	g'u_!||s'ip_!
+_0C	p'Eq_!
 
+// Larger numbers
+_0M0	tS'h-@n       // 10^3
+_1M0	tS'h-@n
+_0M1	m'an          // 10^4
+_1M1	m'an
+_0M2	'@q           // 10^8
+_1M2	'@q
+_0M3	tS;'o         // 10^12
+_1M3	tS;'o
+_0M4	gj'@N
+_1M4	gj'@N
 
 // Character names:
-ㄱ ㄱㅣㅇㅕㄱ
-ㄲ ㅆㅏㅇㄱㅣㅇㅕㄱ
-ㄴ ㄴㅣㅇㅡㄴ
-ㄷ ㄷㅣㄱㅡㄷ
-ㄸ ㅆㅏㅇㄷㅣㄱㅡㄷ
-ㄹ ㄹㅣㅇㅡㄹ
-ㅁ ㅁㅣㅇㅡㅁ
-ㅂ ㅂㅣㅇㅡㅂ
-ㅃ ㅆㅏㅇㅂㅣㅇㅡㅂ
-ㅅ ㅅㅣㅇㅗㅅ
-ㅆ ㅆㅏㅇㅅㅣㅇㅗㅅ
-ㅇ ㅇㅣㅇㅡㅇ
-ㅈ ㅈㅣㅇㅡㅈ
-ㅉ ㅆㅏㅇㅈㅣㅇㅡㅈ
-ㅊ ㅊㅣㅇㅡㅊ
-ㅋ ㅋㅣㅇㅡㅋ
-ㅌ ㅌㅣㅇㅡㅌ
-ㅍ ㅍㅣㅇㅡㅍ
-ㅎ ㅎㅣㅇㅡㅎ
+ᄀ	gij'@q
+ᄁ	's-aNq,ij@q
+ᄂ	ni;'u-n
+ᄃ	diq'u-d-
+ᄄ	's-aNd,iqu-d-
+ᄅ	*i;'u-rr
+ᄆ	mi;'u-m
+ᄇ	pi;'u-p
+ᄈ	's-aNb,i;u-p
+ᄉ	si;'ot-
+ᄊ	's-aNs,i;ot-
+ᄋ	i;'u-N
+ᄌ	tS;i;'u-t
+ᄍ	's-aNdZ;,i;u-t
+ᄎ	tSh-'i;u-t
+ᄏ	khi;'u-k
+ᄐ	thi;'u-t-
+ᄑ	phi;'u-p
+ᄒ	hi;'u-t-
+// end of character names.
+
 
+$textmode
 // Misc:
 // 사ᅵ시ᅩᆺ
 고랫재		고랟째

dictsource/ko_rules

+// eSpeak Korean rules
 // This file is UTF8 encoded
 
+// char names (mostly to catch "ng" sound):
+.group ᄋ
+ᄋ	i;'u-N
+
+
+
+
+
 .replace
 
 // Note: These specific range of "normalized" unicode characters of Korean are
 // h-finals: ㅎ ㄶㅀ
 .L03	ᇂ	ᆭ	ᆶ
 
+// Char names:
+
+
 // Initials
 // Order: ㄱㄲㅋ ㄷㄸㅌ ㅂㅃㅍ ㅅㅆㅈㅉㅊ ㄴㅁ ㄹ ㅎ
 
 		ᆷ		m
 
 .group		ᆼ	// ㅇ
-		ᆼ		nN
+		ᆼ		N
 
 .group		ᇂ	// ㅎ
 		ᇂ		t-

dictsource/pt_list

 _7o	s'EtSim
 _8o	oIt'av
 _9o	n'on
-_10o	d'Esim
+_1Xo	d'Esim
 _2Xo	viZ'Ezim
 _3Xo	tRiZ'Ezim	
 _4Xo	kwad*aZ'Ezim

docs/phonemes.html

 <table>
 <tbody valign=top>
 <tr>
-<td width=25><code>[p]</code><td width=80>
-<td width=25><code>[b]</code><td width=80>
+<td width=25><code>[p]</code><td width=150>
+<td width=25><code>[b]</code><td width=150>
 <tr>
 <td><code>[t]</code><td>
 <td><code>[d]</code><td>
 <table>
 <tbody valign=top>
 <tr><td width=25><code>[@]</code>
-<td width=60>alph<b>a</b><td width=80>schwa
+<td width=60>alph<b>a</b><td width=400>schwa
 
 <tr><td><code>[3]</code>
 <td>bett<b>er</b><td>rhotic schwa. In British English this is the same as <code>[@]</code>, but it includes 'r' colouring in American and other rhotic accents.  In these cases a separate <code>[r]</code> should not be included unless it is followed immediately by another vowel.
 
 <tr><td><code>[a]</code><td>tr<b>a</b>p
 <tr><td><code>[aa]</code><td>b<b>a</b>th<td>This is <code>[a]</code> in some accents, <code>[A:]</code> in others.
-<tr><td><code>[a2]</code><td><b>a</b>bout<td>This may be <code>[@]</code> or may be a more open schwa.
+<tr><td><code>[a#]</code><td><b>a</b>bout<td>This may be <code>[@]</code> or may be a more open schwa.
 <tr><td><code>[A:]</code><td>p<b>al</b>m
 <tr><td><code>[A@]</code><td>st<b>ar</b>t
 <tr><td><p>
 
 </tbody>
 </table>
-
+<p>
+<code> [:] </code> can be used to lengthen a vowel, eg <code> [e:]</code>
 
 </body>
 </html>

phsource/compile_report

                 [l/] fr
 l/l_@           [l/3] base
                 [l/] fr
-l/l@            [h1ø] base
+l/l@            [hÖ{] base
                 [l#] base
                 [l] fr
                 [l/2] fr
 l/L2_uL         [l/2] base
 l/l_3           [l/] de
 l/l_4           [ll] sq
-l/la            [h1ø] base
+l/la            [hÖ{] base
                 [l#] base
                 [l] fr
                 [l/2] fr
                 [K] tn
 l/l_a           [l/3] base
                 [l/] fr
-l/le            [h1ø] base
+l/le            [hÖ{] base
                 [l#] base
                 [l] fr
                 [l/2] fr
                 [&:] af
 l/l_front       [L] sq
 l/l_front_      [l/4] sq
-l/li            [h1ø] base
+l/li            [hÖ{] base
                 [l#] base
                 [l] fr
                 [l/2] fr
 ll/_ll          [L] base
 l/l_long        [l] base
                 [l] fr
-l/lo            [h1ø] base
+l/lo            [hÖ{] base
                 [l#] base
                 [l/2] fr
                 [K] nso
                 [l;] lt
                 [l] ru
                 [l^] ru
-l/lu            [h1ø] base
+l/lu            [hÖ{] base
                 [l#] base
                 [l] fr
                 [l/2] fr
                 [U@] en-n
                 [U@] en-wm
                 [u@] vi
-vdiph2/uw       [u:] en-us
 vdiph2/uw_2     [u:] en
 vdiph2/uw_3     [yU] ro
 vdiph2/uw_4     [u:] en-n
-vdiph2/uw_6     [u#] kk
+vdiph2/uw_6     [u:] en-us
+                [u#] kk
 vdiph2/y@       [y@] zh
 vdiph2/y#@      [Y@] af
 vdiph2/ye       [yE] sq

phsource/ph_english_us

 phoneme u:
   vowel  starttype #u  endtype #u
   length 190
-  FMT(vdiph2/uw)
+  FMT(vdiph2/uw_6)
 endphoneme
 
 

src/numbers.cpp

 }  // end f LookupThousands
 
 
-static int LookupNum2(Translator *tr, int value, int control, char *ph_out)
-{//========================================================================
+static int LookupNum2(Translator *tr, int value, const int control, char *ph_out)
+{//=============================================================================
 // Lookup a 2 digit number
 // control bit 0: ordinal number
 // control bit 1: final tens and units (not number of thousands) (use special form of '1', LANG=de "eins")
 	int ix;
 	int units;
 	int tens;
+	int is_ordinal;
 	int used_and=0;
 	int found_ordinal = 0;
 	int next_phtype;
 		ord_type = 'q';
 	}
 
+	is_ordinal = control & 1;
+
 	if((control & 2) && (n_digit_lookup == 2))
 	{
 		// pronunciation of the final 2 digits has already been found
 			// is there a special pronunciation for this 2-digit number
 			if(control & 8)
 			{
+				// is there a feminine form?
 				sprintf(string,"_%df",value);
 				found = Lookup(tr, string, ph_digits);
 			}
 			else
-			if(control & 1)
+			if(is_ordinal)
 			{
 				strcpy(ph_ordinal, ph_ordinal2);
 		
 
 				if(!found)
 				{
-					sprintf(string,"_%d",value);
-					found = Lookup(tr, string, ph_digits);
+					if((is_ordinal) && (tr->langopts.numbers2 & NUM2_NO_TEEN_ORDINALS))
+					{
+						// don't use numbers 10-99 to make ordinals, always use _1Xo etc (lang=pt)
+					}
+					else
+					{
+						sprintf(string,"_%d",value);
+						found = Lookup(tr, string, ph_digits);
+					}
 				}
 			}
 		}
 			else
 			{
 	
-				if((control & 1) && 
+				if((is_ordinal) && 
 					((units == 0) || (tr->langopts.numbers & NUM_SWAP_TENS) || (tr->langopts.numbers2 & NUM2_MULTIPLE_ORDINAL)))
 				{
 					sprintf(string,"_%dX%c", tens, ord_type);
 							sprintf(string,"_%df",units);
 							found = Lookup(tr, string, ph_digits);
 						}
-						if((control & 1) && ((tr->langopts.numbers & NUM_SWAP_TENS) == 0))
+						if((is_ordinal) && ((tr->langopts.numbers & NUM_SWAP_TENS) == 0))
 						{
 							// ordinal
 							sprintf(string,"_%d%c",units,ord_type);
 			}
 		}
 	
-		if((control & 1) && (found_ordinal == 0) && (ph_ordinal[0] == 0))
+		if((is_ordinal) && (found_ordinal == 0) && (ph_ordinal[0] == 0))
 		{
 			if((value >= 20) && (((value % 10) == 0) || (tr->langopts.numbers & NUM_SWAP_TENS)))
 				Lookup(tr, "_ord20", ph_ordinal);
 		{
 			Lookup(tr, "_0and", ph_and);
 
-			if((control & 1) && (tr->langopts.numbers2 & NUM2_MULTIPLE_ORDINAL))
+			if((is_ordinal) && (tr->langopts.numbers2 & NUM2_MULTIPLE_ORDINAL))
 				ph_and[0] = 0;
 
 			if(tr->langopts.numbers & NUM_SWAP_TENS)
 	int ix;
 	int exact;
 	int ordinal;
+	int tplex;
 	int say_zero_hundred=0;
 	char string[12];  // for looking up entries in **_list
 	char buf1[100];
 			if ((value % 1000) == 0)
 				exact = 1;
 
-			if(LookupThousands(tr, hundreds / 10, thousandplex+1, exact | ordinal, ph_10T) == 0)
+			tplex = thousandplex+1;
+			if(tr->langopts.numbers2 & NUM2_MYRIADS)
+			{
+				tplex = 0;
+			}
+
+			if(LookupThousands(tr, hundreds / 10, tplex, exact | ordinal, ph_10T) == 0)
 			{
 				x = 0;
-				if(tr->langopts.numbers2 & (1 << (thousandplex+1)))
+				if(tr->langopts.numbers2 & (1 << tplex))
 					x = 8;   // use variant (feminine) for before thousands and millions
 				LookupNum2(tr, hundreds/10, x, ph_digits);
 			}
 }  // end of LookupNum3
 
 
+bool CheckThousandsGroup(char *word, int group_len)
+{//================================================
+// Is this a group of 3 digits which looks like a thousands group?
+	int ix;
+
+	if(isdigit(word[group_len]) || isdigit(-1))
+		return(false);
+
+	for(ix=0; ix < group_len; ix++)
+	{
+		if(!isdigit(word[ix]))
+			return(false);
+	}
+	return(true);
+}
+
+
 static int TranslateNumber_1(Translator *tr, char *word, char *ph_out, unsigned int *flags, WORD_TAB *wtab, int control)
 {//=====================================================================================================================
 //  Number translation with various options
 	int decimal_mode;
 	int suffix_ix;
 	int skipwords = 0;
+	int group_len;
 	char *p;
 	char string[32];  // for looking up entries in **_list
 	char buf1[100];
 	n_digits = ix;
 	value = this_value = atoi(word);
 
+	group_len = 3;
+	if(tr->langopts.numbers2 & NUM2_MYRIADS)
+		group_len = 4;
+
 	// is there a previous thousands part (as a previous "word") ?
-	if((n_digits == 3) && (word[-2] == tr->langopts.thousands_sep) && isdigit(word[-3]))
+	if((n_digits == group_len) && (word[-2] == tr->langopts.thousands_sep) && isdigit(word[-3]))
 	{
 		prev_thousands = 1;
 	}
 		// a "thousand"/"million" suffix to this one
 		digix = n_digits + thousands_inc;
 
-		while(((wtab[thousandplex+1].flags & FLAG_MULTIPLE_SPACES) == 0) &&
-			isdigit(word[digix]) && isdigit(word[digix+1]) && isdigit(word[digix+2]) && !isdigit(word[digix+3]) && !isdigit(word[digix-1]))
+		while(((wtab[thousandplex+1].flags & FLAG_MULTIPLE_SPACES) == 0) && CheckThousandsGroup(&word[digix], group_len))
 		{
-			if((word[digix] != '0') || (word[digix+1] != '0') || (word[digix+2] != '0'))
-				thousands_exact = 0;
+			for(ix=0; ix<group_len; ix++)
+			{
+				if(word[digix+ix] != '0')
+				{
+					thousands_exact = 0;
+					break;
+				}
+			}
 
 			thousandplex++;
-			digix += 3;
+			digix += group_len;
 			if((word[digix] == tr->langopts.thousands_sep) || ((tr->langopts.numbers & NUM_ALLOW_SPACE) && (word[digix] == ' ')))
 			{
 				suffix_ix = digix+2;
 	{
 		if(thousands_inc > 0)
 		{
-			if((thousandplex > 0) && (value < 1000))
+			if(thousandplex > 0)
+//			if((thousandplex > 0) && (value < 1000))
 			{
 				if((suppress_null == 0) && (LookupThousands(tr,value,thousandplex, thousands_exact, ph_append)))
 				{

src/synthdata.cpp

 }  // end of SwitchVowelType
 
 
-static int NumInstnWords(USHORT *prog)
-{//===================================
+int NumInstnWords(USHORT *prog)
+{//============================
 	int instn;
 	int instn2;
 	int instn_type;
 	int n;
-	static const char n_words[16] = {1,1,0,0,1,1,1,1,1,2,4,0,0,0,0,0};
+	int type2;
+	static const char n_words[16] = {0,1,0,0,1,1,0,1,1,2,4,0,0,0,0,0};
 
 	instn = *prog;
 	instn_type = instn >> 12;
 	if((n = n_words[instn_type]) > 0)
 		return(n);
 
-	if(instn_type < 4)
+	switch(instn_type)
 	{
+	case 0:
+		if(((instn & 0xf00) >> 8) == i_IPA_NAME)
+		{
+			n = ((instn & 0xff) + 1) / 2;
+			return(n+1);
+		}
+		return(1);;
+
+	case 6:
+		type2 = (instn & 0xf00) >> 9;
+		if((type2 == 5) || (type2 == 6))
+			return(12);  // switch on vowel type
+		return(1);
+
+	case 2:
+	case 3:
 		// a condition, check for a 2-word instruction
 		if(((n = instn & 0x0f00) == 0x600) || (n == 0x0d00))
 			return(2);
 		return(1);
-	}
 
-	// instn_type 11 to 15, 2 words
-	instn2 = prog[2];
-	if((instn2 >> 12) == 0xf)
-	{
-		// This instruction is followed by addWav(), 2 more words
-		return(4);
-	}
-	if(instn2 == i_CONTINUE)
-	{
-		return(3);
+	default:
+		// instn_type 11 to 15, 2 words
+		instn2 = prog[2];
+		if((instn2 >> 12) == 0xf)
+		{
+			// This instruction is followed by addWav(), 2 more words
+			return(4);
+		}
+		if(instn2 == i_CONTINUE)
+		{
+			return(3);
+		}
+		return(2);
 	}
-	return(2);
 }  //  end of NumInstnWords
 
 

src/synthesize.h

 int PauseLength(int pause, int control);
 int LookupPhonemeTable(const char *name);
 unsigned char *GetEnvelope(int index);
+int NumInstnWords(USHORT *prog);
 
 void InitBreath(void);
 

src/tr_languages.cpp

 			tr->langopts.stress_rule = 8;   // ?? 1st syllable if it is heavy, else 2nd syllable
 			tr->langopts.param[LOPT_UNPRONOUNCABLE] = 1;   // disable check for unpronouncable words
 			tr->langopts.numbers = NUM_OMIT_1_HUNDRED;
-			tr->langopts.break_numbers = 0x9999998;
+			tr->langopts.numbers2 = NUM2_MYRIADS;
+			tr->langopts.break_numbers = 0x1111110;
 			tr->langopts.max_digits = 20;
 		}
 		break;
 			tr->langopts.stress_rule = STRESSPOSN_1R;        // stress on final syllable
 			tr->langopts.stress_flags =  0x6 | 0x10 | 0x2000 | 0x20000; 
 			tr->langopts.numbers = NUM_DECIMAL_COMMA | NUM_DFRACTION_2 | NUM_HUNDRED_AND | NUM_AND_UNITS | NUM_ROMAN;
-			tr->langopts.numbers2 = NUM2_MULTIPLE_ORDINAL; 
+			tr->langopts.numbers2 = NUM2_MULTIPLE_ORDINAL | NUM2_NO_TEEN_ORDINALS; 
 			SetLetterVowel(tr,'y');
 			ResetLetterBits(tr,0x2);
 			SetLetterBits(tr,1,"bcdfgjkmnpqstvxz");      // B  hard consonants, excluding h,l,r,w,y

src/translate.h

 	int numbers;
 
 #define NUM2_MULTIPLE_ORDINAL   0x1000
-#define NUM2_ENGLISH_NUMERALS   0x2000
+#define NUM2_NO_TEEN_ORDINALS   0x2000
+#define NUM2_MYRIADS            0x4000
+#define NUM2_ENGLISH_NUMERALS   0x8000
 #define NUM2_THOUSANDS_VAR1     0x40
 #define NUM2_THOUSANDS_VAR2     0x80
 #define NUM2_THOUSANDS_VAR3     0xc0
 	// bits 6-8  use different forms of thousand, million, etc (M MA MB)
 	// bit9=(LANG=rw) say "thousand" and "million" before its number, not after
 	// bit12=(LANG=el,es) use ordinal form of hundreds and tens as well as units
-	// bit13=(LANG=ne)  speak (non-replaced) English numerals in English
+	// bit13=(LANG=pt) don't use 11-19 numbers to make ordinals
+	// bit14=(LANG=ko)  use myriads (groups of 4 digits) not thousands (groups of 3)
+	// bit15=(LANG=ne)  speak (non-replaced) English numerals in English
 	int numbers2;
 
 #define BREAK_THOUSANDS   0x49249248

src/vowelchart.cpp

 	prog_end = prog_start + length;
 	n_vowelfmt_addr = 0;
 
-	for(prog = &phoneme_index[prog_start]; prog < &phoneme_index[prog_end]; prog++)
+	for(prog = &phoneme_index[prog_start]; prog < &phoneme_index[prog_end]; prog += NumInstnWords(prog))
 	{
 		instn = *prog;
 		
-		switch(instn >> 12)
+		if((instn >> 12) == 11)
 		{
-		case 2:
-		case 3:
-			// conditions
-			while((instn & 0xe000) == 0x2000)
-			{
-				instn = *(++prog);
-			}
-			prog--;
-		break;
-
-		case 10:   //  Vowelin, Vowelout
-			prog += 3;
-			break;
-
-		case 9:
-		case 12:   // WAV
-		case 13:   // VowelStart
-		case 14:   // VowelEnd
-		case 15:   // addWav
-			prog++;
-			break;
-
-		case 11:   // FMT
+			// FMT instruction
 			if(n_vowelfmt_addr < N_VOWELFMT_ADDR)
 			{
 				vowelfmt_addr[n_vowelfmt_addr++] = ((instn & 0xf) << 18) + (prog[1] << 2);
 			}
-			prog++;
-
-			break;
 		}
 	}
 }  // end of FindVowelFmt