def filter(ef: ExperimentFiles, overwrite = False):
"""
Filter all of the desired languages for the experiment.
"""
filtration_performed = False
# Iterate over all the languages.
for lang in ef.langs:
orig_path = ef.orig(lang)
filt_path = ef.filtered(lang)
# Don't overwrite already existing files unless
# we've been asked to overwrite.
if not os.path.exists(filt_path) or overwrite:
if not filtration_performed:
REPRO_LOG.log(NORM_LEVEL, "Filtering ODIN data.")
if not USE_CONDOR:
filter_corpus([orig_path], filt_path,
require_aln=True,
require_lang=True,
require_gloss=True,
require_trans=True)
else:
REPRO_LOG.log(NORM_LEVEL, "Filtering {}...".format(lang))
args = ['filter',
'--require-aln',
'--require-lang',
'--require-gloss',
'--require-trans',
orig_path, filt_path]
condorify(args, ef._filter_dir(condor=True), ef.filtered(lang, True))
filtration_performed = True
# If we're using condor, wait until all the
# tasks for this step have completed.
if USE_CONDOR and filtration_performed:
if condor_email:
condor.condor_wait_notify('Filtering of languages performed.', condor_email, "Filtration Done")
else:
condor.condor_wait()
if filtration_performed:
REPRO_LOG.log(NORM_LEVEL, "Filtration complete.")
def extract_pos(ef: ExperimentFiles, overwrite=False):
extraction_performed = False
for tc in ef.tagger_configs():
if not os.path.exists(tc.tagger_path()) or overwrite:
if USE_CONDOR:
args = tc.extract_args(True)
condorify(*args)
else:
args, kwargs = tc.extract_args(False)
extract_from_xigt(args, **kwargs)
extraction_performed = True
if extraction_performed and USE_CONDOR:
if condor_email:
condor.condor_wait_notify("Taggers have been extracted.", condor_email, "CONDOR: POS Tagger Extraction Complete")
else:
condor.condor_wait()
def enrich(ef: ExperimentFiles, overwrite = False, parse=True):
"""
Enrich the files using all types of word alignment, and tag/parse the translation line.
"""
enrichment_performed = False
for lang in ef.langs:
filtered_f = ef.filtered(lang)
enriched_f = ef.enriched(lang)
if not os.path.exists(enriched_f) or overwrite:
# Notify user of enrichment if at least one file is being enriched.
if not enrichment_performed:
REPRO_LOG.log(NORM_LEVEL, "Enriching ODIN data.")
if not USE_CONDOR:
enrich_args = {ALN_VAR:aln_methods,
POS_VAR:[ARG_POS_TRANS],
ARG_INFILE:filtered_f,
ARG_OUTFILE:enriched_f}
intent.commands.enrich.enrich()
else:
args = ['enrich',
'--align', ','.join(aln_methods),
'--pos', ARG_POS_TRANS,
filtered_f, enriched_f]
if parse:
args += ['--parse', 'trans']
condorify(args, ef._enriched_dir(condor=True), ef.enriched(lang, True))
enrichment_performed = True
if USE_CONDOR and enrichment_performed:
if condor_email:
condor.condor_wait_notify('Enrichment of languages performed.', condor_email, "Enrichment Done")
else:
condor.condor_wait()
if enrichment_performed:
REPRO_LOG.log(NORM_LEVEL, "Enrichment complete.")
def eval_taggers_igt(ef: ExperimentFiles, overwrite = False):
evaluation_performed = False
for tc in ef.tagger_configs():
if not os.path.exists(tc.rgigt_eval()) or overwrite:
if not evaluation_performed:
REPRO_LOG.log(NORM_LEVEL, "Evaluating POS taggers on IGT data.")
if USE_CONDOR:
condorify(*tc.eval_igt_pos_args(True))
else:
args, kwargs = tc.eval_igt_pos_args(False)
evaluate_intent(args, **kwargs)
evaluation_performed = True
if USE_CONDOR and evaluation_performed:
if condor_email:
condor.condor_wait_notify("IGT Evaluation Complete", condor_email, subject="IGT POS Evaluation Complete")
else:
condor.condor_wait()
# -------------------------------------------
# Now, consolidate all of the different methods and report.
# -------------------------------------------
te = TaggerEval()
for tc in ef.tagger_configs():
if not os.path.exists(tc.rgigt_eval()):
REPRO_LOG.warn("Missing eval: {}".format(os.path.basename(tc.rgigt_eval())))
else:
with open(tc.rgigt_eval(), 'r') as f:
data = f.readlines()
filename, matches, compares, acc = data[-1].strip().split(',')
te.add_tc(tc, int(matches), int(compares))
print(te)
def postag(ef: ExperimentFiles, overwrite=False):
tagging_performed = False
for tc in ef.tagger_configs():
if not os.path.exists(tc.tagged()) or overwrite:
if USE_CONDOR:
args = tc.tag_args(True)
condorify(*args)
else:
args = tc.tag_args(False)
if tc.method == ARG_POS_CLASS:
intent.commands.enrich.enrich(**args)
else:
do_projection(**args)
tagging_performed = True
if USE_CONDOR and (tagging_performed):
if condor_email:
condor.condor_wait_notify("Data has been tagged.", condor_email, "CONDOR: Tagging Complete")
else:
condor.condor_wait()
for lang in ef.langs:
orig_f = ef.get_original_file(lang)
filtered_f = ef.get_filtered_file(lang)
if not os.path.exists(filtered_f):
filtration_done = True
if USE_CONDOR:
model_prefix, name = ef.get_condor_filter(lang)
run_cmd([p3path, intent_script, 'filter',
'--require-aln', '--require-gloss', '--require-trans', '--require-lang',
orig_f, filtered_f], model_prefix, name, False)
else:
filter_corpus([orig_f], filtered_f, require_lang=True, require_gloss=True, require_trans=True, require_aln=True)
if USE_CONDOR and filtration_done:
condor_wait_notify("Data has been filtered.", email_address, "CONDOR: Filtration complete.")
# -------------------------------------------
# 2) Enriched data
# -------------------------------------------
enrichment_done = False
for lang in ef.langs:
filtered_f = ef.get_filtered_file(lang)
enriched_f = ef.get_enriched_file(lang)
if not os.path.exists(enriched_f):
enrichment_done = True
if USE_CONDOR:
model_prefix, name = ef.get_condor_enrich(lang)
run_cmd([p3path, intent_script, 'enrich',