[erg] _certain_q_rel

Paul Haley paul at haleyai.com
Wed Apr 15 01:53:48 CEST 2015


Your tip on _enough_q_rel had me take a closer look, Dan.

Looks like we copied from the wrong place and that the "ALTKEYREL" 
should be "KEYREL" in the definition we had below.

Thanks,
Paul


On 4/13/2015 5:51 PM, Dan Flickinger wrote:
> I do see the need for an additional analysis of "certain cats" beyond what the trunk ERG provides.  Your guess looks like a good one, to model the new entry on the one for "enough".  But the MRS for "enough cats" comes out looking fine, with no strangeness in the relation for _enough_q_rel, so it is hard to see what would lead to the mangled semantics you reported for the "certain" example.
>
> Anyway, I'll add the missing entry for "certain" to the trunk ERG, and then maybe you can compare it to the one you have proactively added.
>
>   Dan
>
> ----- Original Message -----
> From: "Paul Haley" <paul at haleyai.com>
> To: "Emily M. Bender" <ebender at uw.edu>
> Cc: "erg" <erg at delph-in.net>
> Sent: Monday, April 13, 2015 2:20:56 PM
> Subject: Re: [erg] _certain_q_rel
>
> Dan, you are correct.  Someone here added it, probably because of
> something like this:
>
>
>
> Is certain different than than other quantifiers of the d_-_prt-pl.*_le
> variety?
>
>    * a_great_many_det := d_-_prt-pl_le & [ ORTH < "a", "great", "many" >,
>      SYNSEM [ LKEYS.ALTKEYREL.PRED "_a+great+many_q_rel", PHON.ONSET voc ] ].
>    * enough_det := d_-_prt-plm_le & [ ORTH < "enough" >, SYNSEM [
>      LKEYS.KEYREL.PRED _enough_q_rel, PHON.ONSET voc ] ].
>
> Thanks,
> Paul
>
> On 4/13/2015 4:53 PM, Paul Haley wrote:
>> Good idea, Emily.  After trying to isolate why we couldn't get the
>> on-line demo to demonstrate the same behavior, it appears than an edit
>> to the name of a relation defined in fundamentals.tdl was responsible,
>> although we can't figure out exactly how!
>>
>> It turned out that working through our changes vs. the ERG as is, we
>> found one point at which post-loading of QC.tdl generated some
>> warnings about missing predicates.  Searching that file we cannot find
>> any reference to those predicates, but it is machine-generated in some
>> way.  Could this have led to unifications working (or not) that should
>> (not) have?
>>
>> Just wondering, but thanks for the prod.
>>
>> Paul
>>
>> On 4/9/2015 9:56 PM, Emily M. Bender wrote:
>>> Hi Paul,
>>>
>>> This might be easier to answer if you could send a complete MRS (perhaps
>>> for a shorter example with the same property)...
>>>
>>> Emily
>>>
>>> On Thu, Apr 9, 2015 at 6:24 PM, Paul Haley <paul at haleyai.com
>>> <mailto:paul at haleyai.com>> wrote:
>>>
>>>      Hi Folks,
>>>
>>>      I'm getting the subject relation with non-hole arguments:
>>>
>>>        * {ARG0=e61[declarative, indicative, ¬perfective, ¬progressive,
>>>          untensed]}, {ARG1=x55[individuated, plural, third]}]
>>>
>>>      for a parse of the sentence:
>>>
>>>        * Some organisms survive and others die as the environment
>>>          changes; this changes the percent of organisms with certain
>>>          traits in that population.
>>>
>>>      Here's the syntactic result of that parse, fyi:
>>>
>>>        * 'some'('organisms')('survive')('and'('others'('die'('as'('the'('environment'('changes')))))))('this'('changes'('the'('percent'('of'('organisms'))))('with'('certain'('traits'('in'('that'('population'))))))))
>>>
>>>      The specific lexical entry involved is:
>>>
>>>        * certain_det := d_-_prt-pl_le & [ ORTH < "certain" >, SYNSEM [
>>>          LKEYS.ALTKEYREL.PRED _certain_q_rel, PHON.ONSET con ] ].
>>>
>>>      And the relation is defined as follows:
>>>
>>>        * _certain_q_rel := explicit_quant_agr_q_rel.
>>>
>>>      but I'm stumped on how it's landing up with ARGs instead of
>>>      normal quantification arguments.
>>>
>>>      Does anyone have any thoughts on whether this is proper?
>>>
>>>      Thank you,
>>>      Paul
>>>
>>>
>>>
>>>
>>>
>>>
>>> -- 
>>> Emily M. Bender
>>> Professor, Department of Linguistics
>>> Check out CLMS on facebook! http://www.facebook.com/uwclma
>
>



More information about the erg mailing list