Talk:Subreadings in Constraint Grammar - Revision history

Unhammer at 08:00, 16 May 2013

2013-05-16T08:00:17Z

Unhammer at 07:53, 12 October 2011

2011-10-12T07:53:48Z

Francis Tyers: Created page with '==More discussion==

   +parts are hidden, currently...  Or rather, stuff before the + is hidden. <|krvoje|> o_0  Also, the + is …'

2011-07-11T13:52:23Z

Created page with '==More discussion== <pre> <TinoDidriksen> +parts are hidden, currently... <TinoDidriksen> Or rather, stuff before the + is hidden. <|krvoje|> o_0 <TinoDidriksen> Also, the + is …'

New page

==More discussion==
<pre>

<TinoDidriksen> +parts are hidden, currently...
<TinoDidriksen> Or rather, stuff before the + is hidden.
<|krvoje|> o_0
<TinoDidriksen> Also, the + is not part of the tag.
<TinoDidriksen> +htjeeti becomes a baseform "htjeeti"
<spectie> our thing is a horrible hack :((((
<TinoDidriksen> + is a horrible hack...
<spectie> it's great for apertium, but just doesn't fit into CG yet
<spectie> what cg-proc should do
<spectie> is treat the parts separated by '+' as separate cohorts
<TinoDidriksen> That can relatively easily be done, but sure about that? http://wiki.apertium.org/wiki/Subreadings_in_Constraint_Grammar did not mention that option.
<spectie> the problem is that sometimes we don't want that :))
<spectie> but i think that's a problem with apertium
<TinoDidriksen> Ah
<spectie> we should distinguish + and #
<spectie> + should be for separate cohorts
<spectie> # for separate parts of the same cohort
<spectie> or something
<spectie> e.g. compound words get #
<spectie> and attached clitics get +
<|krvoje|> bbiab, lunch
<spectie> ok
<TinoDidriksen> So, + should be regarded as a soft cohort split thingy...
<spectie> yes
<spectie> but we should also check with unhammer
</pre>

I really think it makes sense to be able to distinguish these two things, we can't use '+' for both:

# Attached clitics / joined words (each word needs to be referred to separately)
# Compound words (we're only interested in the head)

I suggest coming up with some other thing for compounds (perhaps <code>~</code>, although I think we've discussed this before ?

- [[User:Francis Tyers|Francis Tyers]] 13:52, 11 July 2011 (UTC)

@@ Line 40: / Line 40: @@
 - [[User:Francis Tyers|Francis Tyers]] 13:52, 11 July 2011 (UTC)
+==Some file==
+<pre>
+SECTION
+SUBSTITUTE ("од") ("од:5") ("од") (-1 (adj));
+^помладо/adj<pref><comp>+млад<adj><nt><sg><nom><ind>$ ^од/од<pr>$ ^30/30<num>$^./.<sent>$
+</pre>
+<pre>
+MAP (@+FMAINV) TARGET VerbFin ;
+^n'eus/ne<adv>+bezañ<vblex><pri><impers><sp>/ne<adv>+kaout<vblex><pri><p1><pl>$ ^kador/kador<n><f><sg>$ ^ebet/ebet<adv>$^./.<sent>$
+</pre>

@@ Line 13: / Line 13: @@
 <spectie> is treat the parts separated by '+' as separate cohorts
 <TinoDidriksen> That can relatively easily be done, but sure about that? http://wiki.apertium.org/wiki/Subreadings_in_Constraint_Grammar did not mention that option.
+</pre>
+: You can't treat clitics as separate cohorts, how would that even work? Say you have <code>^foo bar/foo<tags>+bar<tags>/foobar<tags>$</code>, should it be treated as "foo<tags>/foobar<tags> followed by a single reading bar<tags>", or  "foo<tags> followed by a bar<tags>/foobar<tags>" ? That sounds like a type of complexity we don't want. [[User:Unhammer|unhammer]] 07:53, 12 October 2011 (UTC)
+<pre>
 <spectie> the problem is that sometimes we don't want that :))
 <spectie> but i think that's a problem with apertium