[erg] a question of parenthesised numbers
Stephan Oepen
oe at ifi.uio.no
Tue Sep 17 11:53:55 CEST 2013
hi david,
how exactly do you invoke PET nowadays?
$ echo "a person (1234543265)" | cheap -sm -repp -cm -default-les=all
-verbose=3 english.grm
[...]
derivation[1] (0):a person (1234543265)
(117 np_frg_c 0 0 3 [root_inffrag]
(116 sp-hd_n_c 0 0 3
(41 a_det/d_-_sg-nmd_le 0 0 1 []
(31 "a" 0 0 1 <0:1>))
(114 hdn-n_prnth_c 0 1 3
(98 n_ms-cnt_ilr 0 1 2
(81 person_n1/n_-_mc_le 0 1 2 []
(30 "person" 0 1 2 <2:8>)))
(105 num_prt-det-nc_c 0 2 3
(95 w_lparen_plr 0 2 3
(93 w_rparen_plr 0 2 3 [w_lparen_plr]
(88 generic_card_ne/aj_-_i-crd-gen_le 0 2 3 [w_rparen_plr
w_lparen_plr]
(36 "(1234543265)" 0 2 3 <9:21>))))))))
there are in total eight analyses, all with the number
as a cardinal adjective. arguably, we should maybe
also allow an analysis as an identifier NE, but that
would cost substantially in ambiguity ...
you can look at tmr/ne{1,2,3}.tdl in the ERG source
for the patterns we recognize through what we call
lightweight named entity recognition.
best, oe
On Tue, Sep 17, 2013 at 9:35 AM, <mottdh at googlemail.com> wrote:
> Hi
>
> i am trying to parse a sentence that has the fragment
>
> a person (1234543265)
>
> I think the ERG treats the parenthetical item as an adjective, which is fine. but the number itself is not recognised, and the parse fails. (it works if i manually add the number to the lexicon!).
>
> my question is: how do i get numbers (and dates by the way) to be recognised as a particular lexical type as if in the lexicon? i am guessing that this is a preparsing issue, and would welcome some pointers to where this might be described.
>
> thank you
>
> David Mott
>
> Sent from my iPad
--
+++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
+++ Universitetet i Oslo (IFI); Boks 1080 Blindern; 0316 Oslo; (+47) 2284 0125
+++ --- oe at ifi.uio.no; stephan at oepen.net; http://www.emmtee.net/oe/ ---
+++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
More information about the erg
mailing list