def __init__(self, mapping, model):
"""
Creates a new AlignIO instance.
@param mapping (Mapping) a mapping table to convert the phone set
"""
# Mapping system for the phonemes
if mapping is None:
mapping = Mapping()
if isinstance(mapping, Mapping) is False:
raise TypeError('Aligner expected a Mapping as argument.')
self._mapping = mapping
# The automatic alignment system
self.aligntrack = AlignTrack( model )
# The list of units file generator
self._listio = ListIO()
# The file names of tracks generator
self._tracknames = TrackNamesGenerator()
class AlignIO:
"""
@author: Brigitte Bigi
@organization: Laboratoire Parole et Langage, Aix-en-Provence, France
@contact: [email protected]
@license: GPL, v3
@copyright: Copyright (C) 2011-2016 Brigitte Bigi
@summary: Read/Write segments from/to Tiers.
??? HOW TO DO: READ ALL ALTERNATIVE LABELS AND MERGE ALTERNATIVE RESULTS ???
"""
def __init__(self, mapping, model):
"""
Creates a new AlignIO instance.
@param mapping (Mapping) a mapping table to convert the phone set
"""
# Mapping system for the phonemes
if mapping is None:
mapping = Mapping()
if isinstance(mapping, Mapping) is False:
raise TypeError('Aligner expected a Mapping as argument.')
self._mapping = mapping
# The automatic alignment system
self.aligntrack = AlignTrack( model )
# The list of units file generator
self._listio = ListIO()
# The file names of tracks generator
self._tracknames = TrackNamesGenerator()
# ------------------------------------------------------------------------
def set_aligner(self, alignername):
"""
Fix the name of the aligner.
The list of accepted aligner names is available in:
>>> aligners.aligner_names()
@param alignername (str - IN) Case-insensitive name of the aligner.
"""
self.aligntrack.set_aligner(alignername)
# -----------------------------------------------------------------------
def get_aligner(self):
"""
Return the aligner name identifier.
"""
return self.aligntrack.get_aligner()
# ----------------------------------------------------------------------
def set_infersp(self, infersp):
"""
Fix the infersp option.
@param infersp (bool - IN) If infersp is set to True, the aligner
will add an optional short pause at the end of each token, and the
will infer if it is relevant.
"""
self.aligntrack.set_infersp( infersp )
# ------------------------------------------------------------------------
def segment_track(self, track, diralign, segment=True):
"""
Perform the speech segmentation of a track in a directory.
@param track (str - int)
@param diralign (str - IN)
@param segment (bool - IN) If True, call an aligner to segment speech,
else create a file with an empty alignment.
@return A message of the aligner in case of any problem, or
an empty string if success.
"""
audiofilename = self._tracknames.audiofilename(diralign, track)
phonname = self._tracknames.phonesfilename(diralign,track)
tokenname = self._tracknames.tokensfilename(diralign,track)
alignname = self._tracknames.alignfilename(diralign,track)
if segment is True:
msg = self.aligntrack.segmenter(audiofilename, phonname, tokenname, alignname)
else:
msg = self.aligntrack.segmenter(audiofilename, None, None, alignname)
return msg
# ------------------------------------------------------------------------
def read(self, dirname):
"""
Read time-aligned tracks in a directory and return a Transcription.
@param diralign (str - IN) Input directory to get files.
@return Transcription
"""
units = self._listio.read( dirname )
trsin = TracksReader()
trsin.set_tracksnames( self._tracknames )
trsin.read(dirname, units)
# map-back phonemes
self._mapping.set_keepmiss( True )
self._mapping.set_reverse( False )
# Map time-aligned phonemes (even the alternatives)
tier = trsin.Find("PhonAlign")
for ann in tier:
for text in ann.GetLabel().GetLabels():
text.SetValue( self._mapping.map_entry(text.GetValue()) )
return trsin
# ------------------------------------------------------------------------
def split(self, inputaudio, phontier, toktier, diralign):
"""
Write tracks from a Transcription.
If the given phontier is not already time-aligned in intervals,
an automatic track-segmenter will be applied first and the TimeAligned
version of the tokenization is returned.
@param audioname (str - IN) Audio file name.
@param phontier (Tier - IN) The phonetization.
@param toktier (Tier - IN) The tokenization, or None.
@param diralign (str - IN) Output directory to store files.
@return Transcription
"""
# Map phonemes from SAMPA to the expected ones.
self._mapping.set_keepmiss( True )
self._mapping.set_reverse( True )
# Map phonetizations (even the alternatives)
for ann in phontier:
for text in ann.GetLabel().GetLabels():
text.SetValue( self._mapping.map( text.GetValue() ) )
sgmt = TrackSplitter()
sgmt.set_tracksnames( self._tracknames )
sgmt.set_trackalign( self.aligntrack )
units = sgmt.split(inputaudio, phontier, toktier, diralign)
self._listio.write(diralign, units)
return sgmt
