From 51cf35538408cf4a2630e0bc77031bd86d6f4351 Mon Sep 17 00:00:00 2001 From: Jacob Date: Sun, 28 Jul 2013 10:57:44 +0100 Subject: sane default settings and more docs --- settings.yaml | 20 +++++++++++--------- 1 file changed, 11 insertions(+), 9 deletions(-) diff --git a/settings.yaml b/settings.yaml index 6bc0e5c..3d0669d 100644 --- a/settings.yaml +++ b/settings.yaml @@ -1,16 +1,18 @@ nbest: 100 # how many entries in the nbest list? corpus: geo # which corpus? [geo, robo] lang: en # which language? [en, de, el, th] -stem: false # run the stemmer? -symm: tgttosrc # which symmetrization? [e.g. srctotgt, tgttosrc, grow, ...] +stem: false # run the stemmer? +symm: srctotgt # which symmetrization? [e.g. srctotgt, tgttosrc, grow, ...] np: true # use NP list? np_type: all # what version of NP list? model: hier # which machine translation model? [phrase, hier] -run: test # which experiment? [dev, test, debug] +run: test # which experiment? [dev, test, debug] workdir: work # where? -filter: false # filter phrase table? -lfrac: 1.0 # what fraction of training sentences should be labeled? -monolingual: false # use monolingual data? -ul_only: false -retrain: false -nlg: true + +# experimental, and unrelated to published work +retrain: false # after tuning, re-extract phrases from tune and train data +filter: false # filter malformed trees from phrase table? +lfrac: 1.0 # what fraction of training sentences should be labeled? +monolingual: false # use monolingual data? +ul_only: false # reweight only with unlabeled data +nlg: false # do MRL->NL rather than semantic parsing -- cgit v1.2.3