diff --git a/src/fst/Makefile.am b/src/fst/Makefile.am
index 3f1f9fc7..bdff92a9 100644
--- a/src/fst/Makefile.am
+++ b/src/fst/Makefile.am
@@ -79,6 +79,139 @@ endif # CAN_FOMA
 #################################################
 #### Add language-specific build rules here: ####
 
+# Hfst - add weights to compounds if using tropical-semiring fst format:
+if WITH_OFST_TROPICAL
+.generated/generator-raw-gt-desc.hfst: .generated/generator-raw-gt-desc.tmp.hfst
+	$(AM_V_REWEIGHT)$(HFST_REWEIGHT) $(MORE_VERBOSITY) $(HFST_FLAGS) \
+			-S '+Cmp' -a 10 --arcs-only -i $< \
+		> $@
+endif
+
+
+
+
+.generated/analyser-gt-norm.%: .generated/analyser-gt-norm.tmp.% \
+					filters/convert_to_flags-CmpNP-tags.% 
+	$(AM_V_XFST_TOOL)$(PRINTF) "read regex \
+			    @\"filters/convert_to_flags-CmpNP-tags.$*\"            \
+			.o. @\"filters/split-CmpNP-tags.hfst\"                     \
+			.o. @\"$<\"                                                \
+			;\n\
+		 twosided flag-diacritics\n\
+		 $(INVERT_HFST)\
+		 save stack $@\n\
+		 quit\n" | $(XFST_TOOL)
+
+# The operation 'twosided flag-diacritics" crashes Xerox badly, so we do with
+# a simpler, less restrictive normative fst when building with Xerox:
+.generated/analyser-gt-norm.xfst: .generated/analyser-gt-norm.tmp.xfst \
+					filters/remove-norm-comp-tags.xfst             
+	$(AM_V_XFST)$(PRINTF) "read regex \
+				@\"filters/remove-norm-comp-tags.xfst\"             \
+			.o. @\"$<\"                                         \
+			;\n\
+		 save stack $@\n\
+		 quit\n" | $(XFST) $(VERBOSITY)
+
+# ... and also in the Oahpa analyser:
+.generated/analyser-oahpa-gt-norm.%: .generated/analyser-oahpa-gt-norm.tmp.% \
+					filters/remove-norm-comp-tags.%             
+	$(AM_V_XFST_TOOL)$(PRINTF) "read regex \
+			    @\"filters/remove-norm-comp-tags.$*\"             \
+			.o. @\"$<\"                                       \
+			;\n\
+		 $(INVERT_HFST)\
+		 save stack $@\n\
+		 quit\n" | $(XFST_TOOL)
+
+# ... and also in the general descriptive analysers:
+.generated/analyser-gt-desc.%: .generated/analyser-gt-desc.tmp.% \
+					filters/remove-norm-comp-tags.%             
+	$(AM_V_XFST_TOOL)$(PRINTF) "read regex \
+			    @\"filters/remove-norm-comp-tags.$*\"             \
+			.o. @\"$<\"                                       \
+			;\n\
+		 $(INVERT_HFST)\
+		 save stack $@\n\
+		 quit\n" | $(XFST_TOOL)
+
+# ... as well as in the HFST disambiguation analyser:
+.generated/analyser-disamb-gt-desc.hfst: .generated/analyser-disamb-gt-desc.tmp.hfst \
+				filters/remove-norm-comp-tags.hfst             
+	$(AM_V_HXFST)$(PRINTF) "read regex \
+		    @\"filters/remove-norm-comp-tags.hfst\"           \
+		.o. @\"$<\"                                           \
+		;\n\
+	invert net\n\
+	save stack $@\n\
+	quit\n" | $(HFST_XFST) -p $(MORE_VERBOSITY)
+
+# ... as well as in the HFST Grammar Checker analyser:
+# (keep the Err/SpaceCmp strings here)
+.generated/analyser-gramcheck-gt-desc.hfst: .generated/analyser-gramcheck-gt-desc.tmp.hfst \
+				filters/remove-norm-comp-tags.hfst           \
+				orthography/comprelax-with-flagtags.compose.hfst
+	$(AM_V_HXFST)$(PRINTF) "\
+	set encode-weights ON\n\
+	read regex \
+		    @\"filters/remove-norm-comp-tags.hfst\"           \
+		.o. @\"$<\"                                           \
+		;\n\
+	twosided flag-diacritics\n\
+	invert net\n\
+	save stack $@\n\
+	quit\n" | $(HFST_XFST) -p $(MORE_VERBOSITY)
+
+# ... as well as in the XEROX/FOMA disambiguation analyser
+# (remove the Err/SpaceCmp thing here):
+.generated/analyser-disamb-gt-desc.%: .generated/analyser-disamb-gt-desc.tmp.% \
+					filters/remove-norm-comp-tags.%             
+	$(AM_V_XFST_TOOL)$(PRINTF) "read regex \
+			    @\"filters/remove-norm-comp-tags.$*\"             \
+			.o. @\"$<\"                                     \
+			;\n\
+		 $(INVERT_HFST)\
+		 save stack $@\n\
+		 quit\n" | $(XFST_TOOL)
+
+# We need to add processing of language-specific tags in the generator:
+define giella_generators
+.generated/generator-gt-%.$(1): .generated/generator-gt-%.tmp.$(1) \
+					 filters/remove-norm-comp-tags.$(1)             
+	$$(AM_V_XFST_TOOL)$$(PRINTF) "read regex \
+			    @\"filters/remove-norm-comp-tags.$(1)\"           \
+			.o. @\"$$<\"                                        \
+			;\n\
+		 $$(INVERT_XFST)$$(INVERT_FOMA)\
+		 save stack $$@\n\
+		 quit\n" | $$(XFST_TOOL)
+endef
+$(foreach fst,hfst xfst foma,$(eval $(call giella_generators,$(fst))))
+
+# We need to add processing of language-specific tags for the Oahpa generators:
+.generated/generator-oahpa-gt-norm.%: .generated/generator-oahpa-gt-norm.tmp.% \
+					filters/remove-norm-comp-tags.%             
+	$(AM_V_XFST_TOOL)$(PRINTF) "read regex \
+			    @\"filters/remove-norm-comp-tags.$*\"             \
+			.o. @\"$<\"                                       \
+			;\n\
+		 $(INVERT_XFST)$(INVERT_FOMA)\
+		 save stack $@\n\
+		 quit\n" | $(XFST_TOOL)
+
+# morpher is a morph segmenting variant: taloautoissani -> talo#auto>i>ssa>ni
+.generated/morpher-gt-desc.%: .generated/morpher-gt-desc.tmp.% \
+					filters/remove-dynamic_compound-char.% \
+					filters/simplify-triple-der-borders.%
+	$(AM_V_XFST_TOOL)$(PRINTF) "read regex \
+			    @\"filters/remove-dynamic_compound-char.$*\".i  \
+			.o. @\"filters/simplify-triple-der-borders.$*\".i  \
+			.o. @\"$<\" \
+			;\n\
+		 $(INVERT_HFST)\
+		 save stack $@\n\
+		 quit\n" | $(XFST_TOOL)
+
 ##################################################################
 #### END: Add local processing instructions ABOVE this line ######
 ##################################################################