labs = io.load_txt_dir(args.txtdir, args.comma_is_pause)
if args.dict == None:
raise SiReError("No path to dictionary. Please use -dict option.")
args.dictionary = dictionary.Dictionary(args.dict[1], args.dict[0])
#The phoneme set used must match the dictionary.
args.phoneme_features = args.dictionary.phoneme_feats
elif args.intype == "hts_lab":
labs = io.open_labdir_line_by_line(args.inpath)
args.intype = "hts_mlf"
elif args.intype == "sire_lab":
labs = io.open_labdir_line_by_line(args.inpath)
else:
if not os.path.exists(args.inpath):
raise SiReError("Input path to mlf does no exist!")
mlf = open(args.inpath, "r").readlines()
labs = io.parse_mlf(mlf, args.intype)
#Used if we make questions fitted to a dataset
#We use qfile as the path to the file later.
if args.questions == True:
parser.add_argument('-qfile', type=str, help="A variable used to store the question file.", default=None)
args.qfile = open(args.qpath, "w")
parser.add_argument('-qfileutt', type=str, help="A variable used to store the GV question file.", default=None)
args.qfileutt = open(args.qpath+"_utt", "w")
for lab in labs:
print "Making full context label for {0}".format(lab[0])
#Make an utt
utt = utterance.Utterance(lab, args)
#This writes out the label and also the questions
write_context_utt(utt, args)
nargs=argparse.REMAINDER,
default='-order 4 -interpolate -gt3min 1 -wbdiscount -debug 3'.split())
parser.add_argument(
'-f',
action='store_true',
help="Force overwrite of outputpath file if it exists.")
parser.add_argument(
'-no_syll_stress',
action='store_true',
help="Replace syllable stress markers with a boundary marker sb.")
args = parser.parse_args()
wf = io.open_writefile_safe(os.path.join(args.outpath, "sents.txt"),
args.f)
labs = io.parse_mlf(io.open_file_line_by_line(args.input_mlf), "align_mlf")
labs = get_phoneme_strings(labs, args.no_syll_stress)
for lab in labs:
wf.write(" ".join(lab) + "\n")
wf.close()
txtpath = os.path.join(args.outpath, "sents.txt")
lmpath = os.path.join(args.outpath, "ngram.lm")
#This allows for people to pass their own options to the ngram binary
options = " " + " ".join(args.lm_binary_options)
subprocess.call(args.ngram_binary_path + " -text " + txtpath + " -lm " +
lmpath + options,
'-collapse_closure',
action="store_true",
help=
"Collapses stops split into closure and release into one when merging state_align_labs with full_context_labs."
)
parser.add_argument('-f',
action="store_true",
help="Force overwrite of files in output dir.")
args = parser.parse_args()
if args.merge_hvite_state_with_sp_align_mlf != None:
state_mlf = sire_io.open_file_line_by_line(
args.merge_hvite_state_with_sp_align_mlf[0])
phone_mlf = sire_io.open_file_line_by_line(
args.merge_hvite_state_with_sp_align_mlf[1])
state_utts = sire_io.parse_mlf(state_mlf, "state_align_mlf")
phone_utts = sire_io.parse_mlf(phone_mlf, "hts_mlf")
merged_utts = merge_hvite_state_with_sp_align_mlf(
state_utts, phone_utts)
if args.f == True:
wf = sire_io.open_writefile_safe(
args.merge_hvite_state_with_sp_align_mlf[2], overwrite=True)
else:
wf = sire_io.open_writefile_safe(
args.merge_hvite_state_with_sp_align_mlf[2])
wf.write("#!MLF!#\n")
for utt in merged_utts:
wf.write("\"*/" + utt.pop(0) + ".rec\"\n")
for phone in utt:
for state in phone:
wf.write(" ".join(state) + "\n")
raise SiReError("No path to dictionary. Please use -dict option.")
args.dictionary = dictionary.Dictionary(args.dict[1], args.dict[0])
#The phoneme set used must match the dictionary.
args.phoneme_features = args.dictionary.phoneme_feats
elif args.intype == "hts_lab":
labs = io.open_labdir_line_by_line(args.inpath)
# print "This is a lab", len(labs[0])
# labs is a list of lists. Each list within the list of one of the labs
args.intype = "hts_mlf"
elif args.intype == "sire_lab":
labs = io.open_labdir_line_by_line(args.inpath)
else:
if not os.path.exists(args.inpath):
raise SiReError("Input path to mlf does no exist!")
mlf = open(args.inpath, "r").readlines()
labs = io.parse_mlf(mlf, args.intype)
#Used if we make questions fitted to a dataset
#We use qfile as the path to the file later.
if args.questions == True:
parser.add_argument('-qfile',
type=str,
help="A variable used to store the question file.",
default=None)
args.qfile = open(args.qpath, "w")
parser.add_argument(
'-qfileutt',
type=str,
help="A variable used to store the GV question file.",
default=None)
args.qfileutt = open(args.qpath + "_utt", "w")
s_lab_count += 1
merged.append(c_merge)
return merged
if __name__ == "__main__":
parser = argparse.ArgumentParser(description='Utility file convertion related methods.')
parser.add_argument('-merge_hvite_state_with_full_context', nargs=3, help="Merge an HVite state level alignment MLF with full-context labels in a directory and output state-level full-context labels to another.", metavar=('mlf_path', 'lab_dir', 'out_dir'))
parser.add_argument('-merge_hvite_state_with_sp_align_mlf', nargs=3, help="Merge an HVite state level alignment MLF which does not contain SP and syllable stress information with a phone level alignment ready mlf which does and output a state-level with SP and syllable stress.", metavar=('state_mlf_path', 'phone_mlf_path', 'out_mlf_path'))
parser.add_argument('-collapse_closure', action="store_true", help="Collapses stops split into closure and release into one when merging state_align_labs with full_context_labs.")
parser.add_argument('-f', action="store_true", help="Force overwrite of files in output dir.")
args = parser.parse_args()
if args.merge_hvite_state_with_sp_align_mlf != None:
state_mlf = sire_io.open_file_line_by_line(args.merge_hvite_state_with_sp_align_mlf[0])
phone_mlf = sire_io.open_file_line_by_line(args.merge_hvite_state_with_sp_align_mlf[1])
state_utts = sire_io.parse_mlf(state_mlf, "state_align_mlf")
phone_utts = sire_io.parse_mlf(phone_mlf, "hts_mlf")
merged_utts = merge_hvite_state_with_sp_align_mlf(state_utts, phone_utts)
if args.f == True:
wf = sire_io.open_writefile_safe(args.merge_hvite_state_with_sp_align_mlf[2], overwrite=True)
else:
wf = sire_io.open_writefile_safe(args.merge_hvite_state_with_sp_align_mlf[2])
wf.write("#!MLF!#\n")
for utt in merged_utts:
wf.write("\"*/"+utt.pop(0)+".rec\"\n")
for phone in utt:
for state in phone:
wf.write(" ".join(state)+"\n")
wf.write(".\n")
wf.close()