def __extractIntentionsFile(self, fn, data_dir, out_fn):
self.logger.info('Making intention level from: %s', fn)
if not os.path.exists(fn):
self.logger.warn('File does not exists, skipping: %s', fn)
return
mlf = ConceptMLF.readFromFile(fn)
fw = MLF()
processor = NadraziProcessor([data_dir])
for mlf_fn, tree in mlf.iteritems():
fw[mlf_fn] = ['%s\n' % processor.process(mlf_fn, tree)]
dirname = os.path.split(fn)[0]
fw.writeToFile(os.path.join(dirname, out_fn))
def __extractIntentionsFile(self, fn, data_dir, out_fn):
self.logger.info('Making intention level from: %s', fn)
if not os.path.exists(fn):
self.logger.warn('File does not exists, skipping: %s', fn)
return
mlf = ConceptMLF.readFromFile(fn)
fw = MLF()
processor = NadraziProcessor([data_dir])
for mlf_fn, tree in mlf.iteritems():
fw[mlf_fn] = ['%s\n' % processor.process(mlf_fn, tree)]
dirname = os.path.split(fn)[0]
fw.writeToFile(os.path.join(dirname, out_fn))
def mlfparametrize(self,
mlf,
inDir=None,
outDir=None,
source_type='normalized',
target_type='decoded',
use_empty=False):
self.setupDirs(inDir, outDir)
foo, new_sen_file = tempfile.mkstemp()
self.logger.info("Reading MLF")
mlf = MLF.readFromFile(mlf)
self.logger.info("Making build directories")
self.makeDirs()
self.logger.info("Getting text from XMLs")
if new_sen_file is None:
fn = os.path.join(self.settings['PDT_WORK_DIR'], 'sen_file')
else:
fn = new_sen_file
fw = codecs.open(fn, 'w', 'utf-8')
try:
for fn in self.getInputXMLs():
dxml = DXML.readFromFile(fn)
acts = dxml.getDialogueActs(source_type)
idx = 0
for utter in acts:
for txt, attrs in utter:
if not txt:
self.logger.info(
"Empty <dialogue_act> in file: %s", fn)
key = "*/%s_%05d" % (os.path.splitext(
os.path.basename(fn))[0], idx)
if key in mlf:
line = mlf[key]
elif not use_empty:
line = [txt + '\n']
else:
line = ['\n']
line = '\n'.join(line)
fw.write(line)
idx += 1
finally:
fw.close()
self.parametrize(target_type, new_sen_file)
os.remove(new_sen_file)
def mlfparametrize(self, mlf, inDir=None, outDir=None,
source_type='normalized', target_type='decoded', use_empty=False):
self.setupDirs(inDir, outDir)
foo, new_sen_file = tempfile.mkstemp()
self.logger.info("Reading MLF")
mlf = MLF.readFromFile(mlf)
self.logger.info("Making build directories")
self.makeDirs()
self.logger.info("Getting text from XMLs")
if new_sen_file is None:
fn = os.path.join(self.settings['PDT_WORK_DIR'], 'sen_file')
else:
fn = new_sen_file
fw = codecs.open(fn, 'w', 'utf-8')
try:
for fn in self.getInputXMLs():
dxml = DXML.readFromFile(fn)
acts = dxml.getDialogueActs(source_type)
idx = 0
for utter in acts:
for txt, attrs in utter:
if not txt:
self.logger.info("Empty <dialogue_act> in file: %s", fn)
key = "*/%s_%05d" % (os.path.splitext(os.path.basename(fn))[0], idx)
if key in mlf:
line = mlf[key]
elif not use_empty:
line = [txt + '\n']
else:
line = ['\n']
line = '\n'.join(line)
fw.write(line)
idx += 1
finally:
fw.close()
self.parametrize(target_type, new_sen_file)
os.remove(new_sen_file)
def main(
self,
model_dir,
encoding=None,
batch=False,
omit_leaves=False,
mlf=False,
xml_dir=None,
ref_mlf=None,
skip_empty=False,
input_chain=None,
batch_size=100,
no_underscores=True,
force_pdt=False,
pdt_dir=None,
):
encoding = sys.stdout.encoding
if encoding is None:
if os.name == "nt":
encoding = "cp1250"
else:
encoding = "iso-8859-2"
datasets_fn = pjoin(model_dir, "datasets")
datasets_fr = file(datasets_fn, "r")
datasets = []
isymMaps = []
for i, line in enumerate(datasets_fr):
line = line.strip()
datasets.append(line)
if line != "off":
isymMaps.append(SymMap.readFromFile(pjoin(model_dir, "isym%d.map" % (i + 1,))))
osymMap = SymMap.readFromFile(pjoin(model_dir, "osym.map"))
if "signed" in datasets:
da_type = "signed"
else:
da_type = "normalized"
if not pdt_dir:
pdt_dir = "/opt/PDT-2.0/tools/machine-annotation"
if xml_dir:
reader = input.MultiReader([xml_dir], input.DXMLReader)
if force_pdt and "lemma" in datasets or "pos" in datasets:
if os.name == "nt":
raise RuntimeError("Datasets 'lemma' and 'pos' are unsupported on Windows")
reader = input.PDTReader(pdt_dir, reader, online=not batch)
else:
reader = input.StdInReader(encoding=encoding, type=da_type)
if "lemma" in datasets or "pos" in datasets:
if os.name == "nt":
raise RuntimeError("Datasets 'lemma' and 'pos' are unsupported on Windows")
reader = input.PDTReader(pdt_dir, reader, online=not batch)
if input_chain is not None:
reader = input.InputChain(input_chain, reader)
generator = input.InputGenerator(reader, datasets, datasets[0], noUnderscores=no_underscores)
hypMLF = MLF()
refMLF = MLF()
if not batch:
for da_fn, da_id, da_semantics, da_txts in generator.readInputs():
da_empty = not bool(da_semantics.strip())
if da_empty and skip_empty:
continue
refMLF[da_id] = da_semantics + "\n"
dcd = self.parseLine(model_dir, [da_txts], isymMaps, osymMap, omitLeaves=omit_leaves)
if dcd:
if len(dcd) == 1:
hypMLF[da_id] = dcd[0].encode(encoding) + "\n"
else:
hypMLF[da_id] = ";".join(dcd).encode(encoding) + "\n"
else:
hypMLF[da_id] = line + "\n"
if not mlf:
print hypMLF[da_id],
else:
all_processed = False
inputs = generator.readInputs()
while not all_processed:
da_count = 0
lines = []
ids = []
for da_fn, da_id, da_semantics, da_txts in inputs:
da_empty = not bool(da_semantics.strip())
if da_empty and skip_empty:
continue
refMLF[da_id] = da_semantics + "\n"
lines.append(da_txts)
ids.append(da_id)
da_count += 1
if da_count >= batch_size:
break
else:
all_processed = True
dcd = self.parseLine(model_dir, lines, isymMaps, osymMap, omitLeaves=omit_leaves)
for da_id, ol in zip(ids, dcd):
hypMLF[da_id] = ol.encode(encoding) + "\n"
if not mlf:
print hypMLF[da_id],
if mlf:
s = "".join(hypMLF.toLines())
print s
if ref_mlf:
refMLF.writeToFile(ref_mlf)
def main(self,
model_dir,
encoding=None,
batch=False,
omit_leaves=False,
mlf=False,
xml_dir=None,
ref_mlf=None,
skip_empty=False,
input_chain=None,
batch_size=100,
no_underscores=True,
force_pdt=False,
pdt_dir=None):
encoding = sys.stdout.encoding
if encoding is None:
if os.name == 'nt':
encoding = 'cp1250'
else:
encoding = 'iso-8859-2'
datasets_fn = pjoin(model_dir, 'datasets')
datasets_fr = file(datasets_fn, 'r')
datasets = []
isymMaps = []
for i, line in enumerate(datasets_fr):
line = line.strip()
datasets.append(line)
if line != 'off':
isymMaps.append(
SymMap.readFromFile(
pjoin(model_dir, 'isym%d.map' % (i + 1, ))))
osymMap = SymMap.readFromFile(pjoin(model_dir, 'osym.map'))
if 'signed' in datasets:
da_type = 'signed'
else:
da_type = 'normalized'
if not pdt_dir:
pdt_dir = '/opt/PDT-2.0/tools/machine-annotation'
if xml_dir:
reader = input.MultiReader([xml_dir], input.DXMLReader)
if force_pdt and 'lemma' in datasets or 'pos' in datasets:
if os.name == 'nt':
raise RuntimeError(
"Datasets 'lemma' and 'pos' are unsupported on Windows"
)
reader = input.PDTReader(pdt_dir, reader, online=not batch)
else:
reader = input.StdInReader(encoding=encoding, type=da_type)
if 'lemma' in datasets or 'pos' in datasets:
if os.name == 'nt':
raise RuntimeError(
"Datasets 'lemma' and 'pos' are unsupported on Windows"
)
reader = input.PDTReader(pdt_dir, reader, online=not batch)
if input_chain is not None:
reader = input.InputChain(input_chain, reader)
generator = input.InputGenerator(reader,
datasets,
datasets[0],
noUnderscores=no_underscores)
hypMLF = MLF()
refMLF = MLF()
if not batch:
for da_fn, da_id, da_semantics, da_txts in generator.readInputs():
da_empty = not bool(da_semantics.strip())
if (da_empty and skip_empty):
continue
refMLF[da_id] = da_semantics + '\n'
dcd = self.parseLine(model_dir, [da_txts],
isymMaps,
osymMap,
omitLeaves=omit_leaves)
if dcd:
if len(dcd) == 1:
hypMLF[da_id] = dcd[0].encode(encoding) + '\n'
else:
hypMLF[da_id] = ';'.join(dcd).encode(encoding) + '\n'
else:
hypMLF[da_id] = line + '\n'
if not mlf:
print hypMLF[da_id],
else:
all_processed = False
inputs = generator.readInputs()
while not all_processed:
da_count = 0
lines = []
ids = []
for da_fn, da_id, da_semantics, da_txts in inputs:
da_empty = not bool(da_semantics.strip())
if (da_empty and skip_empty):
continue
refMLF[da_id] = da_semantics + '\n'
lines.append(da_txts)
ids.append(da_id)
da_count += 1
if da_count >= batch_size:
break
else:
all_processed = True
dcd = self.parseLine(model_dir,
lines,
isymMaps,
osymMap,
omitLeaves=omit_leaves)
for da_id, ol in zip(ids, dcd):
hypMLF[da_id] = ol.encode(encoding) + '\n'
if not mlf:
print hypMLF[da_id],
if mlf:
s = ''.join(hypMLF.toLines())
print s
if ref_mlf:
refMLF.writeToFile(ref_mlf)
