[logon] reflexives

[Francis Bond]

G'day,

2008/11/20 Stephan Oepen <oe at ifi.uio.no>:
> hi again, francis,
>
>  [...]
>
>> myself/ourselves/yourself/yourselves/himselg/herself/itself/theirselves.
>>
>> Is this also the case in noen?
>>
>> I tried an enhanced version:
>>
>> reflexive_pronoun_ef := monotonic_mtr &
>> [ INPUT.RELS < [ PRED pron_rel, LBL #h0, ARG0 #x1 ],
>>                [ PRED identity_rel, LBL #h0, ARG0 #x0, ARG1 #x1 ] >,
>>   FILTER.RELS < [ PRED poss_rel, ARG2 #x1 ] >,
>>   OUTPUT.RELS < +copy+ & [ ARG0 [ PRONTYPE refl ] ],
>>                 [ PRED num_equate, ARG0 #x1, ARG1 #x0 ],
>>                 [ PRED gend_equate, ARG0 #x1, ARG1 #x0 ] > ].
>>
>> identity_ditch_ef := elision_mtr &
>> [ INPUT.RELS < [ PRED identity_rel ] > ].
>>
>> but still no luck.  I will try to debug more on the morrow.
>
> you helped me recover some of the more intricate LOGON memories :-): i
> think you are right that grammatically `bound' pronouns should always
> pick up NUM and GEND from their antecedents.  in principle, the above
> is in the right spirit, but doing this as an ERG accomodation rule is
> too late in this case.  the NUM and GEND `equate' machinery is part of
> the (grammar-independent) snugging phase already.
>
> i just checked in a revision to the Transfer Matrix and NoEn, adding a
> pair of rules (in `uio/snug.mtr'):
>
>  identity_equate_num_cf := monotonic_mtr &
>  [ CONTEXT.RELS < [ PRED identity_rel, LBL #h0, ARG0 #x0, ARG1 #x1 ],
>                   [ PRED pron_rel, LBL #h0, ARG0 #x1 & x ] >,
>    FILTER.RELS < [ PRED num_equate, ARG0 #x1, ARG1 #x0 ] >,
>    OUTPUT.RELS < [ PRED num_equate, ARG0 #x1, ARG1 #x0 ] >,
>    FLAGS.EQUAL < #x1 > ].
>
>  identity_equate_gend_cf := monotonic_mtr &
>  [ CONTEXT.RELS < [ PRED identity_rel, LBL #h0, ARG0 #x0, ARG1 #x1 ],
>                   [ PRED pron_rel, LBL #h0, ARG0 #x1 & x ] >,
>    FILTER.RELS < [ PRED gend_equate, ARG0 #x1, ARG1 #x0 ] >,
>    OUTPUT.RELS < [ PRED gend_equate, ARG0 #x1, ARG1 #x0 ] >,
>    FLAGS.EQUAL < #x1 > ].
>
> i also moved the original `reflexive_pronoun_ef' into `uio/erg.mtr' and
> had to rationalize the use of `poss' in NoEn in order to be able to do
> that.  but all for the better, i am sure!
>
> hence, in principle, `identity' should now behave the way you expected
> it to in the first place, without any additional special rules in JaEn.
> would you have the time to test this claim for us?

If the input is fairly fully specified (only really true for pronouns
in Japanese) then this worked fine.  To make it work for less
specified input, I added rules to
(a) equate person (making the default be 3)
(b) specialize number (no_number in the target forking to sq,sq and pl,pl)

This works unless the input is a zero pronoun, which doesn't get
generated as we were looking for a "u" and it has become an "x".
This means we have to generate the zero pronouns before the num_equate
rules.   I will try some more jiggling.

> --- we should really start writing up some of that snugging magic, i.e.
> the various special predicates supported only transfer-internally ...

Gosh, wouldn't a description of all this make a great book (or even
better wiki :-).

-- 
Francis Bond <http://www2.nict.go.jp/x/x161/en/member/bond/>
NICT Language Infrastructure Group