[developers] Question on using PET/ACE for parsing

Johnny Wei jwei at umass.edu
Wed Jun 13 21:09:00 CEST 2018


Hi Prof. Oepen,

I have tried using the command such as 'echo 'this is a _generic_vbd_ .' |
./bin/cheap -default-les=all -cm lingo/erg/english.grm' on the top level of
logon, but have had no luck. Cheap outputs that it did not find any lexicon
entries for '_generic_vbd_'.

One thing that I am noticing is that if I look in the _generic_vbd_ entry
in the linguistic type database (link
<http://compling.hss.ntu.edu.sg/ltdb/cgi/ERG_1214//showtype.cgi?typ=v_np%2A_pa-unk_le>)
and choose any surface form listed to replace the word in the sentence, it
will activate the generic_vbd rule, and not any other rule. Could there be
any potential downsides to using this method?

Again, I really appreciate your time! Thanks!

On Tue, Jun 12, 2018 at 5:35 PM, Stephan Oepen <oe at ifi.uio.no> wrote:

> hmm, what if you disable PoS tagging, i.e. drop the ‘-tagger’ option to
> cheap?
>
> oe
>
>
> On Tue, 12 Jun 2018 at 23:23 Johnny Wei <jwei at umass.edu> wrote:
>
>> Hi,
>>
>> Thank you for your response! Is there anyway to prevent sentences like
>> "_generic_vbd_ is a person." to be parseable where "_generic_vbd_" is being
>> recognized as a "NN" POS generic? Would this require turning off the
>> unknown word machinery and adding these generics as lexical entries?
>>
>> I am interested because the grammaticality of my language model is
>> important to me. Thanks! Please let me know!
>>
>>
>> On Tue, Jun 12, 2018 at 2:10 PM, Stephan Oepen <oe at ifi.uio.no> wrote:
>>
>>> hi johnny,
>>>
>>> > [...] A sample from my language model might look like
>>> "generic_proper_ne had
>>> > VBP_u_unknown a cat ." I want to see if these sequences can be parsed
>>> by the
>>> > ERG.
>>>
>>> if you look at ‘gle.tdl’ in the ERG sources, it provides the
>>> definitions of the generic lexical entries that are put to use in
>>> unknown word handling.  for debugging purposes (similar to what you
>>> have in mind, i think) they all have a unique orthography—which you
>>> should be able to just give to the parser; please see the attachment,
>>> for how this works on the ERG on-line interface.
>>>
>>> best wishes, oe
>>>
>>
>>
>>
>> --
>> Johnny Wei
>>
>


-- 
Johnny Wei
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.delph-in.net/archives/developers/attachments/20180613/5ed7da5f/attachment.html>


More information about the developers mailing list