New language option lowercaseSentence

When set in a language file, a period followed by a lower case letter is
detected as end of sentence. Normal behavior is to require a capital
letter.

.gitignore

@@ -118,6 +118,7 @@ tests/*.check
 !tests/cmd_options.test
 !tests/windows-data.test
 !tests/windows-installer.test
+!tests/voices.test
 
 espeak-ng.pc
 

CHANGELOG.md

@@ -13,6 +13,7 @@ The espeak-ng project is a fork of the espeak project.
 *  Add more tests to check the various parts of espeak-ng.
 *  Various changes to clean up the codebase.
 *  Restructure "option brackets" language option to "brackets" and "bracketsAnnounced".
+*  New Language option: "lowercaseSentence" for ending a sentence if a period is followed by a lower case letter.
 *  Added voice variants
 
 documentation:
@@ -29,6 +30,7 @@ updated languages:
 *  grc (Ancient Greek) -- Reece Dunn (support for variant Greek letter forms)
 *  hak (Hakka Chinese) -- Juho Hiltunen
 *  haw (Hawaiian) -- Juho Hiltunen
+*  ka (Georgian) -- Juho Hiltunen
 *  kok (Konkani) -- Juho Hiltunen
 *  nb (Norwegian Bokmål) -- Juho Hiltunen
 *  nci (Classical Nahuatl) -- Juho Hiltunen

Makefile.am

@@ -275,6 +275,7 @@ check:	tests/encoding.check \
 	tests/variants.check \
 	tests/windows-installer.check \
 	tests/bom.check \
+	tests/voices.check \
 	tests/cmd_options.check
 
 ##### fuzzer:

docs/voices.md

@@ -26,6 +26,7 @@
   - [phonemes](#phonemes)
   - [dictionary](#dictionary)
   - [dictrules](#dictrules)
+  - [lowercaseSentence](#lowercaseSentence)
   - [replace](#replace)
   - [stressRule](#stressrule)
   - [stressLength](#stresslength)
@@ -385,6 +386,16 @@ language dictionary. They apply to rules in the language's `*_rules`
 dictionary file and also its `*_list` exceptions list. See
 [Text to Phoneme Translation](dictionary.md#conditional-rules).
 
+### lowercaseSentence
+
+	lowercaseSentence <no arguments>
+
+By default, a sentence end is detected if a period `.` is followed by an uppercase letter.
+When lowercaseSentence is set, a period followed by a lowercase letter is also handled as end of sentence.
+
+Note that other conditions, such as abbreviations, might override this setting.
+
+
 ### replace
 
 	replace <flags> <phoneme> <replacement phoneme>

espeak-ng-data/lang/ccs/ka

@@ -1,2 +1,3 @@
 name Georgian
 language ka
+lowercaseSentence	// A period followed by a lowercase letter is considered a sentence

src/libespeak-ng/readclause.c

@@ -876,9 +876,9 @@ int ReadClause(Translator *tr, char *buf, short *charix, int *charix_top, int n_
 							else if (iswlower(c_next) || (c_next == '-')) // hyphen is needed for lang-hu (eg. 2.-kal)
 								is_end_clause = false; // only if followed by lower-case, (or if there is a XML tag)
 						} 
-						if (iswlower(c_next)) {
+						if (iswlower(c_next) && tr->langopts.lowercase_sentence == false) {
 							// next word has no capital letter, this dot is probably from an abbreviation
-							is_end_clause = 0;
+							is_end_clause = false;
 						}
 						if (any_alnum == false) {
 							// no letters or digits yet, so probably not a sentence terminator

src/libespeak-ng/tr_languages.c

@@ -295,6 +295,7 @@ static Translator *NewTranslator(void)
 	tr->langopts.replace_chars = NULL;
 	tr->langopts.alt_alphabet_lang = L('e', 'n');
 	tr->langopts.roman_suffix = utf8_null;
+	tr->langopts.lowercase_sentence = false;
 
 	SetLengthMods(tr, 201);
 

src/libespeak-ng/translate.h

@@ -540,6 +540,7 @@ typedef struct {
 	int max_lengthmod;
 	int lengthen_tonic;   // lengthen the tonic syllable
 	int suffix_add_e;      // replace a suffix (which has the SUFX_E flag) with this character
+	bool lowercase_sentence;	// when true, a period . causes a sentence stop even if next character is lowercase
 } LANGUAGE_OPTIONS;
 
 typedef struct {
@@ -608,6 +609,7 @@ typedef struct {
 	int end_stressed_vowel;  // word ends with stressed vowel
 	int prev_dict_flags[2];     // dictionary flags from previous word
 	int clause_terminator;
+
 } Translator;
 
 #define OPTION_EMPHASIZE_ALLCAPS  0x100

src/libespeak-ng/voices.c

@@ -92,6 +92,7 @@ enum {
 	V_BREATHW,
 
 	// these override defaults set by the translator
+	V_LOWERCASE_SENTENCE,
 	V_WORDGAP,
 	V_INTONATION,
 	V_TUNES,
@@ -122,6 +123,8 @@ static MNEM_TAB keyword_tab[] = {
 	{ "maintainer",   V_MAINTAINER },
 	{ "status",       V_STATUS },
 
+
+	{ "lowercaseSentence",	V_LOWERCASE_SENTENCE },
 	{ "variants",     V_VARIANTS },
 	{ "formant",      V_FORMANT },
 	{ "pitch",        V_PITCH },
@@ -654,6 +657,14 @@ voice_t *LoadVoice(const char *vname, int control)
 		case V_FORMANT:
 			VoiceFormant(p);
 			break;
+		case V_LOWERCASE_SENTENCE: {
+			if (langopts)
+				langopts->lowercase_sentence = true;
+			else
+				fprintf(stderr, "Cannot set lowercaseSentence: language not set, or is invalid.\n");
+			break;
+			}
+
 		case V_PITCH:
 			// default is  pitch 82 118
 			if (sscanf(p, "%d %d", &pitch1, &pitch2) == 2) {

tests/voices.test

@@ -0,0 +1,27 @@
+#!/bin/sh
+
+test_voices() {
+	TEST_LANG=$1
+	MESSAGE=$2
+	EXPECTED=$3
+	TEST_TEXT=$4
+	TEST_BROKEN=${5:-false}
+
+	if [ "x$MESSAGE" = x- ] ; then
+		echo "testing ${TEST_LANG}"
+	else
+		echo "testing ${TEST_LANG} ($MESSAGE)"
+	fi
+	ESPEAK_DATA_PATH=`pwd` LD_LIBRARY_PATH=src:${LD_LIBRARY_PATH} \
+		src/espeak-ng ${OPTIONS} -xq -v ${TEST_LANG} "${TEST_TEXT}" > actual.txt
+	echo "${EXPECTED}" > expected.txt
+	if [ x$TEST_BROKEN = xbroken ] ; then
+		diff expected.txt actual.txt || (echo "... ignoring error (broken)" && true)
+	else
+		diff expected.txt actual.txt || exit 1
+	fi
+}
+
+##### Voice options
+test_voices ka "lowercaseSentence" "s'asi,amovn,oa t#k#v'eni g'ats#noba
+r'ogor brdz'andebit#" "სასიამოვნოა თქვენი გაცნობა. როგორ ბრძანდებით" #872