[developers] Edge can be built interactively, but isn't in the chart

[Emily M. Bender]

Yes -- that is one of the known cases. However, it's not what's going on
here: The daughters of the missing edge can be used to create the analogous
edge in the sentences with the same verb twice. (And in one case, the
daughter is already the product of a syntax rule.)

Thank you for the guess though! I'm hoping some such guess will put me on
the right path...

Emily

On Mon, Feb 24, 2020 at 10:34 AM Stephan Oepen <oe at ifi.uio.no> wrote:

> from memory, i believe the chart display shows edges with a non-empty
> orthographemic todo list, i.e. a remaining need to pass through a lexical
> rule with an associated orthographemic effect.  this property of edges is
> not visible in the interface, and interactive unification may not be paying
> attention to it.  upon completion of lexical parsing, only edges with an
> empty todo list can go on and feed into syntax rules, so this filter that
> is applied by the parser might explain seeming misalignment between the
> interactive mode and what actually happens during parsing.
>
> just a wild guess :-), oe
>
>
> On Mon, 24 Feb 2020 at 18:49 Emily M. Bender <ebender at uw.edu> wrote:
>
>> Dear all,
>>
>> [Cross-posted to developers and the delphinqa.]
>>
>> After 16 years of teaching grammar engineering, I thought I'd found all
>> of the ways in which one can be in the situation of seemingly being able to
>> build an edge through interactive unification which isn't in the chart.
>> I've documented all of the ones I know about here:
>>
>> http://moin.delph-in.net/GeFaqUnifySurprise
>>
>> Alas, I've found evidence of a new one. Or rather: I'm in that situation
>> (together with a student) but none of the cases noted there apply. More
>> specifically, with the grammar for Meithei [mni] that can be found here:
>>
>> http://faculty.washington.edu/ebender/mni-debug.tgz
>>
>> If we try to analyze this sentence:
>>
>> yoŋ-siŋ túm-í čá-í
>> monkey-PL sleep-NHYP eat-NHYP
>> Monkeys sleep and eat.
>>
>> The LKB and ace both return no parses found. If instead of using the two
>> verbs (one intransitive and one transitive but with a dropped object), we
>> repeat either one of the verbs, we get the expected parses (with both the
>> LKB and ace).
>>
>> yoŋ-siŋ čá-í čá-í
>> yoŋ-siŋ túm-í túm-í
>>
>> Returning to the non-parsing sentence, and looking at the LKB's parse
>> chart, what's missing is the VP-T built out of applying the VP1-TOP-COORD
>> rule to the VP-B over čá-í and the VP over túm-í.  Puzzlingly, I can build
>> this edge interactively just fine. I've run out of guesses as to why it's
>> not showing up in the char and so I thought I'd put this puzzle out in case
>> other DELPH-INites might be entertained by it.
>>
>> Curiously,
>> Emily
>>
>> p.s. Discourse directed me to an earlier discussion about this, where
>> @johnca suggested checking that *chart-packing-p* is set to NIL. It is.
>>
>>
>> --
>> Emily M. Bender (she/her)
>> Howard and Frances Nostrand Endowed Professor
>> Department of Linguistics
>> Faculty Director, CLMS
>> University of Washington
>> Twitter: @emilymbender
>>
>

-- 
Emily M. Bender (she/her)
Howard and Frances Nostrand Endowed Professor
Department of Linguistics
Faculty Director, CLMS
University of Washington
Twitter: @emilymbender
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.delph-in.net/archives/developers/attachments/20200224/1b15a269/attachment-0001.html>