Work with urj-Cyrl generation

and filterning
giellalt · Aug 28, 2024 · 1a1c1ec · 1a1c1ec
1 parent d214d90
commit 1a1c1ec
Show file tree

Hide file tree

Showing 3 changed files with 68 additions and 18 deletions.
diff --git a/src/fst/morphology/Makefile.modifications-local.am b/src/fst/morphology/Makefile.modifications-local.am
@@ -19,9 +19,14 @@
 
 if HAVE_SHARED_URJ_CYRL
 generated_files/urj-Cyrl-$(GLANG)-propernouns.lexc: \
-	$(gt_SHARED_urj_Cyrl)/src/fst/stems/urj-Cyrl-propernouns.lexc
+		$(gt_SHARED_urj_Cyrl)/src/fst/stems/urj-Cyrl-propernouns.lexc \
+		$(top_srcdir)/src/fst/scripts/urj-$(GTLANG)-conversion.pl
 	$(AM_V_at)$(MKDIR_P) generated_files
-	$(AM_V_GEN)cp -f $< $@
+	$(AM_V_GEN)rm -f $@ && \
+	echo "" >> $@ && \
+	echo "! <--- Dump from URJ-CYRL -->" >> $@ && \
+	echo "" >> $@ && \
+	$(top_srcdir)/src/fst/scripts/urj-$(GTLANG)-conversion.pl $< >> $@
 else
 generated_files/mul-$(GLANG)-%.lexc:
 	$(AM_V_at)$(MKDIR_P) generated_files

diff --git a/src/fst/morphology/affixes/propernouns.lexc b/src/fst/morphology/affixes/propernouns.lexc
@@ -161,7 +161,7 @@ LEXICON CYRL-A_SUR !!= * @CODE@
 ! +Sem/Sur: N-ava_01 ;
 ! +Sem/Sur: ENDLEX ;
 
-LEXICON PROP_KIT_SUR !!= * @CODE@
+LEXICON PROP_KAL_SUR !!= * @CODE@
  +Sem/Sur: N-ava_01 ;
  +Sem/Sur: ENDLEX ;
 
@@ -177,9 +177,9 @@ LEXICON PROP_LAK_MAL !!= * @CODE@
  +Sem/Mal: N-ava_01 ;
  +Sem/Mal: ENDLEX ;
 
-LEXICON PROP_KIT_MAL !!= * @CODE@
- +Sem/Mal: N-ava_01 ;
- +Sem/Mal: ENDLEX ;
+!LEXICON PROP_KIT_MAL !!= * @CODE@
+! +Sem/Mal: N-ava_01 ;
+! +Sem/Mal: ENDLEX ;
 
 LEXICON PROP_OSH_MAL !!= * @CODE@
  +Sem/Mal: N-ava_01 ;
@@ -214,9 +214,9 @@ LEXICON PROP_KAL_FEM !!= * @CODE@
 !LEXICON PROP_LAK_FEM
 ! +Sem/Fem: N-ava_01 ;
 
-LEXICON PROP_KIT_FEM !!= * @CODE@
- +Sem/Fem: N-ava_01 ;
- +Sem/Fem: ENDLEX ;
+!LEXICON PROP_KIT_FEM !!= * @CODE@
+! +Sem/Fem: N-ava_01 ;
+! +Sem/Fem: ENDLEX ;
 
 LEXICON PROP_OSH_FEM !!= * @CODE@
  +Sem/Fem: N-ava_01 ;
@@ -242,20 +242,20 @@ LEXICON LEXC_PROP_KUDO_FEM !!= * @CODE@
 
 !! PLACE NAMES FROM TEMPLATE 
 
-LEXICON PROP-PLC_KAL !!= * @CODE@
- +Sem/Plc: N-ava_01 ;
- +Sem/Plc: ENDLEX ;
+!LEXICON PROP-PLC_KAL !!= * @CODE@
+! +Sem/Plc: N-ava_01 ;
+! +Sem/Plc: ENDLEX ;
 
-LEXICON PROP-PLC_KIT !!= * @CODE@
- +Sem/Plc: N-ava_01 ;
- +Sem/Plc: ENDLEX ;
+!LEXICON PROP-PLC_KIT !!= * @CODE@
+! +Sem/Plc: N-ava_01 ;
+! +Sem/Plc: ENDLEX ;
 
 !LEXICON PROP-PLC_OSH
 ! +Sem/Plc: N-ava_01 ;
 
-LEXICON PROP-PLC_KUDO !!= * @CODE@
- +Sem/Plc: N-ava_01 ;
- +Sem/Plc: ENDLEX ;
+!LEXICON PROP-PLC_KUDO !!= * @CODE@
+! +Sem/Plc: N-ava_01 ;
+! +Sem/Plc: ENDLEX ;
 
 
 

diff --git a/src/fst/scripts/urj-mhr-conversion.pl b/src/fst/scripts/urj-mhr-conversion.pl
@@ -0,0 +1,45 @@
+#!/usr/bin/perl -w
+#
+# urj-mhr-conversion.pl
+# Convert names in the URJ propernoun lexicon to mhr.
+
+use strict;
+use utf8;
+use feature 'unicode_strings';
+BEGIN {
+       $| = 1;
+       binmode(STDIN, ':encoding(UTF-8)');
+       binmode(STDOUT, ':encoding(UTF-8)');
+}
+use open qw( :encoding(UTF-8) :std );
+
+while(<>) {
+	# Continuation lexicon substitutions:
+#	s/ C-FI-NEN/nen LONDON/g ;
+	s/ PROP_KAL_PLC/ PROP-PLC_/g ;
+	s/ PROP_KUDO_PLC/ PROP-PLC_/g ;
+	s/ PROP_OSH_PLC/ PROP-PLC_/g ;
+	s/ PROP-PLC_KEL1/ PROP-PLC_/g ;
+	s/ PROP-PLC_VELE/ PROP-PLC_/g ;
+	s/ PROP-PLC_KUDO/ PROP-PLC_/g ;
+	s/ PROP_RUS_JA/ PROP-PLC_/g ;
+	s/ PROP_KUDO / PROP_ /g ;
+	s/ PROP_KAL / PROP_ /g ;
+	s/ PROP_OSH / PROP_ /g ;
+
+
+# loanwords with compound border over identical vowels
+	s/Hjarteelva/Hjarte-elva/g ;
+
+
+# names with Inari Saami inflection
+	s/^Aanaar\+/!Aanaar+/g ;
+
+# sme special symbols
+	s/b9/b/g ;
+
+	my $line = $_;
+
+	print $line;
+}
+