diff options
Diffstat (limited to 'scripts/ems/experiment.meta')
-rw-r--r-- | scripts/ems/experiment.meta | 44 |
1 files changed, 32 insertions, 12 deletions
diff --git a/scripts/ems/experiment.meta b/scripts/ems/experiment.meta index 9ce378a1a..57ef4f9d6 100644 --- a/scripts/ems/experiment.meta +++ b/scripts/ems/experiment.meta @@ -7,8 +7,15 @@ get-corpus default-name: corpus/txt rerun-on-change: input-extension output-extension template: IN OUT $input-extension $output-extension +pre-tok-clean + in: raw-stem + out: pre-tok-cleaned + default-name: corpus/pre-tok-cleaned + pass-unless: pre-tok-clean + template: $pre-tok-clean IN $input-extension $output-extension OUT OUT.lines-retained + parallelizable: yes tokenize - in: raw-stem + in: pre-tok-cleaned out: tokenized-stem default-name: corpus/tok pass-unless: input-tokenizer output-tokenizer @@ -158,11 +165,18 @@ get-corpus pass-unless: get-corpus-script default-name: lm/txt template: $get-corpus-script > OUT +use-parallel-corpus + in: parallel-corpus-stem + out: tokenized-corpus + default-name: lm/tok + ignore-unless: parallel-corpus-stem + template: ln -s IN.$output-extension OUT tokenize in: raw-corpus out: tokenized-corpus default-name: lm/tok pass-unless: output-tokenizer + ignore-if: parallel-corpus-stem template: $output-tokenizer < IN > OUT parallelizable: yes mock-parse @@ -185,7 +199,7 @@ lowercase default-name: lm/lowercased pass-unless: output-lowercaser ignore-if: output-truecaser - only-factor-0: yes + #only-factor-0: yes template: $output-lowercaser < IN > OUT parallelizable: yes truecase @@ -204,8 +218,14 @@ split default-name: lm/split pass-unless: output-splitter template: $output-splitter -model IN1.$output-extension < IN > OUT +strip + in: split-corpus + out: stripped-corpus + default-name: lm/stripped + pass-unless: mock-output-parser-lm + template: $moses-script-dir/training/strip-xml.perl < IN > OUT train - in: split-corpus + in: stripped-corpus out: lm default-name: lm/lm ignore-if: rlm-training @@ -220,7 +240,7 @@ randomize pass-unless: lm-randomizer ignore-if: rlm-training train-randomized - in: split-corpus + in: stripped-corpus out: rlm default-name: lm/rlm ignore-unless: rlm-training @@ -953,21 +973,21 @@ split-reference-devtest ignore-unless: use-mira multiref: $moses-script-dir/ems/support/run-command-on-multiple-refsets.perl template: $output-splitter -model IN1.$output-extension < IN > OUT -reduce-reference +strip-reference in: split-ref out: reference - default-name: tuning/reference.reduced + default-name: tuning/reference.stripped pass-unless: mock-output-parser-references multiref: $moses-script-dir/ems/support/run-command-on-multiple-refsets.perl - template: $moses-script-dir/training/reduce-factors.perl --factor 0 --xml 1 --corpus IN --reduced-corpus OUT && $moses-script-dir/training/wrappers/mosesxml2brackets.py < IN > OUT.trees -reduce-reference-devtest + template: $moses-script-dir/training/strip-xml.perl < IN > OUT && $moses-script-dir/training/wrappers/mosesxml2brackets.py < IN > OUT.trees +strip-reference-devtest in: split-ref-devtest out: reference - default-name: tuning/reference.devtest.reduced + default-name: tuning/reference.devtest.stripped pass-unless: mock-output-parser-references ignore-unless: use-mira multiref: $moses-script-dir/ems/support/run-command-on-multiple-refsets.perl - template: $moses-script-dir/training/reduce-factors.perl --factor 0 --xml 1 --corpus IN --reduced-corpus OUT && $moses-script-dir/training/wrappers/mosesxml2brackets.py < IN > OUT.trees + template: $moses-script-dir/training/strip-xml.perl < IN > OUT && $moses-script-dir/training/wrappers/mosesxml2brackets.py < IN > OUT.trees filter in: input TRAINING:sigtest-filter-phrase-translation-table TRAINING:sigtest-filter-reordering-table TRAINING:corpus-mml-prefilter=OR=TRAINING:corpus-mml-postfilter=OR=TRAINING:domains TRAINING:transliteration-table out: filtered-dir @@ -1224,13 +1244,13 @@ lowercase-reference pass-if: recaser multiref: $moses-script-dir/ems/support/run-command-on-multiple-refsets.perl template: $output-lowercaser < IN > OUT -reduce-reference +strip-reference in: lowercased-reference out: reference default-name: evaluation/reference pass-unless: mock-output-parser-references multiref: $moses-script-dir/ems/support/run-command-on-multiple-refsets.perl - template: $moses-script-dir/training/reduce-factors.perl --factor 0 --xml 1 --corpus IN --reduced-corpus OUT && $moses-script-dir/training/wrappers/mosesxml2brackets.py < IN > OUT.trees + template: $moses-script-dir/training/strip-xml.perl < IN > OUT && $moses-script-dir/training/wrappers/mosesxml2brackets.py < IN > OUT.trees wade in: filtered-dir truecased-input tokenized-reference alignment system-output out: wade-analysis |