[developers] Options to extract syntax trees from FFTB

[Stephan Oepen]

hi tuấn anh,

from what i recall about how FFTB writes tsdb(1) profiles, this should be
easy: once treebanking is complete, the ‘result’ relation should contain
one entry per item for each active derivation, typically one after full
disambiguation.

the ‘derivation’ field will always be there, but i am not quite sure
whether FFTB writes the ‘tree’ (labeled phrase structure) and ‘mrs’ fields?
 you should be able to observe that in your profiles.

if not, the LOGON ‘redwoods’ script can recreate labeled trees for each
derivation, using a command roughly like the following:

$LOGONROOT/redwoods —terg —export tree —target /tmp <profile>

best wishes, oe


On Thu, 27 Feb 2020 at 08:48 Tuấn Anh Lê <tuananh.ke at gmail.com> wrote:

> Hi everyone,
>
> We are trying to use FFTB to tree bank a small corpus and we would like to
> extract the chosen syntax trees from the corpus. The expected output would
> be something like
>
> It works --> ("S" ("NP" ("NP" ("it"))) ("VP" ("V" ("V" ("works")))))
>
> Is there a way to extract this from the FFTB profile?
>
> Currently I'm selecting the trees manually by parsing the sentences using
> ACE with the options --report-label and then split the output string with "
> ;  " but I'm not sure if this is the best approach.
>
> [erg-trunk]$ ace -g erg-0.9.26.dat --report-label
> It works
> SENT: It works
> [ LTOP: h0 INDEX: e2 [ e SF: prop TENSE: pres MOOD: indicative PROG: -
> PERF: - ] RELS: < [ pron<0:2> LBL: h4 ARG0: x3 [ x PERS: 3 NUM: sg GEND: n
> PT: std ] ]  [ pronoun_q<0:2> LBL: h5 ARG0: x3 RSTR: h6 BODY: h7 ]  [
> _work_v_1<3:8> LBL: h1 ARG0: e2 ARG1: x3 ARG2: i8 ] > HCONS: < h0 qeq h1 h6
> qeq h4 > ICONS: < > ] ;  ("S" ("NP" ("NP" ("it"))) ("VP" ("V" ("V"
> ("works")))))
> NOTE: 1 readings, added 391 / 68 edges to chart (27 fully instantiated, 35
> actives used, 18 passives used) RAM: 1880k
>
> Thank you
> --
> Tuan Anh
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.delph-in.net/archives/developers/attachments/20200227/85c694ad/attachment.html>