Replacement for flag diacritics

From Apertium
Revision as of 15:48, 14 June 2014 by Francis Tyers (talk | contribs)
Jump to navigation Jump to search

People like to use flag diacritics for stuff. But they are bad because they are ugly and get in the way of stuff.

Alternative: Use symbols and finite-state operations!

We have < and > for morphological tags, and { and } for archiphonemes and morphological features. We add a new type of symbol with [ and ] for modelling morphotactic restrictions.

Example

Multichar_Symbols

%<v%> %<cop%> %<tv%> %<aor%> %<prog%> %<p1% %<p3%> %<sg%>

%[%-aor%] %[%+aor%]

%+

LEXICON Root

Verbs ; 

LEXICON PERS

%<p1%>%<sg%>:im # ;
%<p3%>%<sg%>: # ;

LEXICON COP

%+i%<cop%>%<aor%>%[%+aor%]: PERS ;

LEXICON V-TV 

%<v%>%<tv%>%<aor%>%[%+aor%]:ar PERS ;
%<v%>%<tv%>%<aor%>%[%+aor%]:ar COP ;
%<v%>%<tv%>%<prog%>%[%-aor%]:iyor COP ;

LEXICON Verbs 

bil:bil V-TV ; ! ""
Alphabet

b i l m i y o r u m 

%<v%> %<tv%> %<prog%> %<aor%> %<p1%> %<p2%> %<p3%> %<sg%> %<cop%>

%[%+aor%]:0  %[%-aor%]:0 

;

Sets 

Verb = %<v%> ;

Rules 

"No consecutive +aor tags"
%[%+aor%]:0 /<= %[%+aor%]:0 :* _ ; 
$ hfst-lexc test.lexc | hfst-invert -o test.hfst
$ hfst-twolc test-const.twol -o const.hfst
$ hfst-compose-intersect -1 test.hfst -2 const.hfst | hfst-fst2strings 

biliyorim:bil<v><tv><prog>+i<cop><aor><p1><sg>
biliyor:bil<v><tv><prog>+i<cop><aor><p3><sg>
bilarim:bil<v><tv><aor><p1><sg>
bilar:bil<v><tv><aor><p3><sg>