Armenian emphasis mark (U+055B) is used for interjections, so treat it as an exclamation mark.

src/libespeak-ng/tokenizer.c

@@ -38,7 +38,6 @@
 static const unsigned short punct_chars[] = {
 	0x00a1, // inverted exclamation
 	0x00bf, // inverted question
-	0x055b, // Armenian emphasis mark
 	0
 };
 
@@ -46,7 +45,6 @@ static const unsigned short punct_chars[] = {
 static const unsigned int punct_attributes[] = {
 	CLAUSE_SEMICOLON | CLAUSE_OPTIONAL_SPACE_AFTER,  // inverted exclamation
 	CLAUSE_SEMICOLON | CLAUSE_OPTIONAL_SPACE_AFTER,  // inverted question
-	CLAUSE_PERIOD | CLAUSE_PUNCTUATION_IN_WORD,  // Armenian emphasis mark
 	0
 };
 

tests/tokenizer.c

@@ -66,7 +66,7 @@ test_armenian()
 {
 	printf("testing Armenian (Armn) script classification\n");
 
-	assert(clause_type_from_codepoint(0x055B) == (CLAUSE_PERIOD | CLAUSE_PUNCTUATION_IN_WORD));
+	assert(clause_type_from_codepoint(0x055B) == (CLAUSE_EXCLAMATION | CLAUSE_PUNCTUATION_IN_WORD));
 	assert(clause_type_from_codepoint(0x055C) == (CLAUSE_EXCLAMATION | CLAUSE_PUNCTUATION_IN_WORD));
 	assert(clause_type_from_codepoint(0x055D) == CLAUSE_COMMA);
 	assert(clause_type_from_codepoint(0x055E) == (CLAUSE_QUESTION | CLAUSE_PUNCTUATION_IN_WORD));