def fill_gaps(tier, mark):
'''For a given tier, add an interval with a given mark to all unfilled intervals'''
old_time = tier.minTime
for interval in tier.intervals:
if old_time < interval.minTime:
tier.addInterval(textgrid.Interval(old_time, interval.minTime, mark))
old_time = interval.maxTime
if tier.maxTime is not None and old_time < tier.maxTime:
tier.addInterval(textgrid.Interval(old_time, tier.maxTime, mark))
def overwrite_interval(int_tier, start, stop, mark):
'''Given an interval, add it to the interval
overwriting anything within that time span '''
min_ind = int_tier.indexContaining(start)
max_ind = int_tier.indexContaining(stop)
#Get rid of anything in between
del int_tier.intervals[min_ind + 1:max_ind]
min_ind = int_tier.indexContaining(start)
max_ind = int_tier.indexContaining(stop)
minOverlap = int_tier[min_ind]
maxOverlap = int_tier[max_ind]
if minOverlap == maxOverlap:
if stop != minOverlap.maxTime:
newOver = textgrid.Interval(stop, minOverlap.maxTime,
minOverlap.mark)
minOverlap.maxTime = start
int_tier.addInterval(newOver)
else:
minOverlap.maxTime = sp.minTime
else:
minOverlap.maxTime = start
maxOverlap.minTime = stop
int_tier.add(start, stop, mark)
transcr_ph_txt = endpoints[5]
offsets = decoded_results["output"][0]["offsets"]
txg.endTime = offsets[len(offsets) - 1]
words_tier = textgrid.IntervalTier("words", 0,
offsets[len(offsets) - 1])
phones_tier = textgrid.IntervalTier("phones", 0,
offsets[len(offsets) - 1])
for jx in range(0, len(transcr_wrd_txt) - 1):
#print(transcr_wrd_txt[jx]+" "+str(transcr_wrd_starts[jx])+" "+str(transcr_wrd_ends[jx]))
#if transcr_wrd_starts[jx] == transcr_wrd_ends[jx]: transcr_wrd_ends[jx] += 0.01
words_tier.addInterval(
textgrid.Interval(transcr_wrd_starts[jx], transcr_wrd_ends[jx],
transcr_wrd_txt[jx]))
for px in range(0, len(transcr_ph_txt) - 1):
#print(transcr_ph_txt[px]+" "+str(transcr_ph_starts[px])+" "+str(transcr_ph_ends[px]))
#if transcr_ph_starts[jx] == transcr_ph_ends[jx]: transcr_ph_ends[jx] += 0.01
#if transcr_ph_txt[px] == " ": transcr_ph_txt[px] = "SIL"
phones_tier.addInterval(
textgrid.Interval(transcr_ph_starts[px], transcr_ph_ends[px],
transcr_ph_txt[px]))
#print(transcr[ix] + " " + str(offs[ix]))
txg.append(words_tier)
txg.append(phones_tier)
txg.write(args.outputfile)
import textgrid
f = textgrid.TextGrid()
f.read("A_man_bought_a_car_in_Paris-DC-5.TextGrid")
print(f.getNames())
print(f.getFirst("stress"))
for y in f.getList("stress"):
for x in y:
print(x)
print("Duration:", x.duration())
print("Bounds:", x.bounds())
print("Overlaps:", x.overlaps(textgrid.Interval(0.0, 1.89568, "a")))
romajiTier.name = path
else:
romajiTier.name = os.path.splitext(name)[0]
# convert kanji to romaji
for interval in Tier.intervals:
# skip intervals with dummy silence
if interval.mark == "#":
continue
# take interval information from original
# and apply convertor to the interval name
else:
romajiInt = textgrid.Interval(
minTime=interval.minTime,
maxTime=interval.maxTime,
mark=conv.do(interval.mark))
# convert long chounpu with the preceding vowel
romajiInt.mark = re.sub("([a-z])\s*-", "\1\1",
romajiInt.mark)
romajiTier.addInterval(romajiInt)
# create an empty TextGrid object to write romaji tier to
newTg = textgrid.TextGrid()
newTg.append(romajiTier)
# write textgrid to file
print("saving to {}".format(os.path.join(args.outputDir, name)))
newTg.write(os.path.join(args.outputDir, name))