[developers] Latest ERG not generating unknown names with ACE

[Michael Wayne Goodman]

This change to lfr.tdl looked promising (`svn diff -r22903:25380 lfr.tdl`):

 proper_ne+name_lfr := lexical_filtering_rule &
 [ +CONTEXT < [ SYNSEM [ PHON.ONSET con_or_voc,
-                        LOCAL [ CAT.HEAD noun,
+                        LOCAL [ CAT [ HEAD noun,
+              VAL.COMPS *olist* ],
                                 CONJ cnil ],
-                        LKEYS.KEYREL.PRED abstr_named_rel ] ] >,
+                        LKEYS.KEYREL.PRED abstr_named_rel ],
+               ORTH.LD #ld ] >,
   +INPUT < [ SYNSEM [ PHON.ONSET unk_onset,
-                      LKEYS.KEYREL.PRED named_rel ] ] >,
+                      LKEYS.KEYREL.PRED named_rel ],
+             ORTH.LD #ld ] >,
   +OUTPUT  < >,
   +POSITION "I1 at C1" ].

But just reverting this change was not enough:

$ echo "Ubuntu is dropping Unity." | ace -g erg.dat -Tq1 2>/dev/null | ace
-g erg.dat -e 2>/dev/null
Unity, _generic_proper_ne_ is dropping.
_generic_proper_ne_ is dropping unity.

Are there any other relevant files/types regarding generic named entities
and generation?


On Fri, Mar 10, 2017 at 4:54 PM, Francis Bond <bond at ieee.org> wrote:

> G'day,
>
> we have found what looks like a regression in the newest ERG.
>
> For my tried and trusty ERG (trunk: Revision: 22903) I can parse and the
> generate an unknown name:
>
> echo "Kuririnrinrin!" | ace -g /home/bond/svn/erg/erg.dat -Tf | ace -e -g
> /home/bond/svn/erg/erg.dat
> NOTE: 1 readings, added 78 / 4 edges to chart (4 fully instantiated, 3
> actives used, 2 passives used) RAM: 285k
> NOTE: parsed 1 / 1 sentences, avg 285k, time 0.00556s
> Kuririnrinrin.
> NOTE: 15 passive, 163 active edges in final generation chart; built 15
> passives total. [1 results]
> NOTE: generated 1 / 1 sentences, avg 432k, time 0.00410s
>
> For a newer ERG (Revision: 25380), I just get "_generic_proper_ne_":
>
> echo "Kuririnrinrin!" | ace -g ~/grammars/erg/erg-2017-03-10.dat -T | ace
> -g ~/grammars/erg/erg-2017-03-10.dat -e
> NOTE: 1 readings, added 79 / 4 edges to chart (5 fully instantiated, 2
> actives used, 2 passives used) RAM: 213k
> NOTE: parsed 1 / 1 sentences, avg 213k, time 0.00320s
> _generic_proper_ne_.
> NOTE: 16 passive, 168 active edges in final generation chart; built 16
> passives total. [3 results]
> NOTE: generated 1 / 1 sentences, avg 491k, time 0.00555s
>
> ace is the same version in both cases (0.9.24).
>
> Any idea how to regain the old behavior?
>
> --
> Francis Bond <http://www3.ntu.edu.sg/home/fcbond/>
> Division of Linguistics and Multilingual Studies
> Nanyang Technological University
>



-- 
Michael Wayne Goodman
Ph.D. Candidate, UW Linguistics
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.delph-in.net/archives/developers/attachments/20170414/3409fc1c/attachment.html>