def generate_elan(elan, audio):
for j in range(80):
i = j + 1
if os.path.isfile(elan + 'new_elan' + str(i) + '.eaf'):
count = 0
tree = pr(elan + 'new_elan' + str(i) + '.eaf')
annotationTree = tree.documentElement
headers = annotationTree.getElementsByTagName("HEADER")
for header in headers:
media = header.getElementsByTagName("MEDIA_DESCRIPTOR")
for med in media:
if count == 0:
if med.hasAttribute(
"MEDIA_URL"
) and not med.hasAttribute("EXTRACTED_FROM"):
count = 1
name = "file:///" + audio + str(i) + ".mp4"
med.setAttribute("MEDIA_URL", name)
if med.hasAttribute(
"RELATIVE_MEDIA_URL"
) and not med.hasAttribute("EXTRACTED_FROM"):
count = 1
name = "./" + str(i) + ".mp4"
med.setAttribute("RELATIVE_MEDIA_URL", name)
else:
if med.hasAttribute("EXTRACTED_FROM"):
name = "file:///" + audio + str(i) + ".wav"
med.setAttribute("EXTRACTED_FROM", name)
if med.hasAttribute("MEDIA_URL"):
name = "file:///" + audio + str(i) + ".wav"
med.setAttribute("MEDIA_URL", name)
if med.hasAttribute("RELATIVE_MEDIA_URL"):
name = "./" + str(i) + ".wav"
med.setAttribute("RELATIVE_MEDIA_URL", name)
open(elan + 'elan' + str(i) + '.eaf', 'w').close()
new_efname = elan + 'elan' + str(i) + '.eaf'
tree.writexml(open(new_efname, 'w'),
indent=" ",
addindent=" ",
newl='\n')
def getElanTime(efname):
tree = pr(efname)
annotationTree = tree.documentElement
timeslot = None
timeslot2 = None
t = None
t2 = None
#go to the notes tier
tiers = annotationTree.getElementsByTagName("TIER")
for tier in tiers:
if tier.getAttribute("TIER_ID") == "default": #"notes":
annotations = tier.getElementsByTagName("ANNOTATION")
for annotation in annotations:
align_annotations = annotation.getElementsByTagName(
"ALIGNABLE_ANNOTATION")
for align_annotation in align_annotations:
annotation_value = align_annotation.getElementsByTagName(
"ANNOTATION_VALUE")[0]
if "start_Being_Heard_one" in annotation_value.childNodes[
0].data:
timeslot = align_annotation.getAttribute(
"TIME_SLOT_REF1")
if "start_Being_Heard_two" in annotation_value.childNodes[
0].data:
timeslot2 = align_annotation.getAttribute(
"TIME_SLOT_REF1")
print("Found start_Being_Heard_one annotation at ", timeslot)
if timeslot2 is not None:
print("Found start_Being_Heard_two annotation at ", timeslot2)
#go to the time tier
times = annotationTree.getElementsByTagName("TIME_ORDER")
for time in times:
time_slots = time.getElementsByTagName("TIME_SLOT")
for time_slot in time_slots:
if time_slot.getAttribute("TIME_SLOT_ID") == timeslot:
t = time_slot.getAttribute("TIME_VALUE")
if time_slot.getAttribute("TIME_SLOT_ID") == timeslot2:
t2 = time_slot.getAttribute("TIME_VALUE")
print("Elan annotation times - ", t, "", t2)
return t, t2
def generate_elan(dirc):
tree = pr('elan.eaf')
annotationTree = tree.documentElement
for j in range(80):
i = j + 1
audiofile = dirc + str(i) + ".wav"
if os.path.isfile(audiofile):
headers = annotationTree.getElementsByTagName("HEADER")
for header in headers:
media = header.getElementsByTagName("MEDIA_DESCRIPTOR")
for med in media:
if med.hasAttribute("MEDIA_URL") and not med.hasAttribute(
"EXTRACTED_FROM"):
name = "file:///" + dirc + str(i) + ".mp4"
med.setAttribute("MEDIA_URL", name)
if med.hasAttribute(
"RELATIVE_MEDIA_URL"
) and not med.hasAttribute("EXTRACTED_FROM"):
name = "./" + str(i) + ".mp4"
med.setAttribute("RELATIVE_MEDIA_URL", name)
if med.hasAttribute("EXTRACTED_FROM"):
name = "file:///" + dirc + str(i) + ".mp4"
med.setAttribute("EXTRACTED_FROM", name)
if med.hasAttribute("MEDIA_URL") and med.hasAttribute(
"EXTRACTED_FROM"):
name = "file:///" + dirc + str(i) + ".wav"
med.setAttribute("MEDIA_URL", name)
if med.hasAttribute(
"RELATIVE_MEDIA_URL") and med.hasAttribute(
"EXTRACTED_FROM"):
name = "./" + str(i) + ".wav"
med.setAttribute("RELATIVE_MEDIA_URL", name)
new_efname = dirc + 'elan' + str(i) + '.eaf'
tree.writexml(open(new_efname, 'w'),
indent=" ",
addindent=" ",
newl='\n')
def fixElanOffest(efname, dirc, offset):
tree = pr(dirc + "/" + efname)
annotationTree = tree.documentElement
#go to the time tier
times = annotationTree.getElementsByTagName("TIME_ORDER")
for time in times:
time_slots = time.getElementsByTagName("TIME_SLOT")
for time_slot in time_slots:
if int(time_slot.getAttribute("TIME_VALUE")) + offset > 0:
time_slot.setAttribute(
"TIME_VALUE",
str(int(time_slot.getAttribute("TIME_VALUE")) + offset))
new_efname = dirc + "/" + 'new_' + efname
tree.writexml(open(new_efname, 'w'),
indent=" ",
addindent=" ",
newl='\n')
print("Fixed offfset output at ", new_efname)
return
def main(args):
""" parse command like argument"""
parser = argparse.ArgumentParser()
parser.add_argument("-b", "--block")
parser.add_argument("-f", "--file", nargs ='*')
parser.add_argument("-m", "--facet", nargs = '*')
parser.add_argument("-s", "--start", nargs ='*')
parser.add_argument("-e", "--end", nargs ='*')
parser.add_argument('--combine', dest='combine', action='store_true')
parser.add_argument('--no-combine', dest='combine', action='store_false')
parser.set_defaults(combine=True)
parser.add_argument('--weak', dest='weak', action='store_true')
parser.add_argument('--no-weak', dest='weak', action='store_false')
parser.set_defaults(weak=False)
parser.add_argument('--all', dest='all', action='store_true')
parser.add_argument('--only-human', dest='all', action='store_false')
parser.set_defaults(all=True)
parser.add_argument('--mean-normalize', dest='mean', action='store_true')
parser.add_argument('--no-mean-normalize', dest='mean', action='store_false')
parser.set_defaults(mean=True)
parser.add_argument('--global-mean', dest='global_mean', action='store_true')
parser.add_argument('--no-global-mean', dest='global_mean', action='store_false')
parser.set_defaults(mean=False)
parser.add_argument('--binary', dest='binary', action='store_true')
parser.add_argument('--no-binary', dest='binary', action='store_false')
parser.set_defaults(mean=False)
args = parser.parse_args()
if args.block is not None:
block = int(args.block)
else:
block = 1000
filename = args.file
print(filename)
facet_file = args.facet
combine = args.combine
weak = args.weak
all = args.all
mean = args.mean
global_mean = args.global_mean
binary = args.binary
start_time_list = args.start if args.start is not None else None
end_time_list = args.end if args.end is not None else None
print("file =", filename, "facet file =", facet_file, "block =", block, "combine =", combine, "weak =", weak)
time_dict_list = []
emotion_kappa = []
total = len(filename)
for i in range(total):
print(filename[i])
tree = pr(filename[i])
annotationTree = tree.documentElement
tiers = annotationTree.getElementsByTagName("TIER")
#put time tier to a dictionary
ts_dict = {}
times = annotationTree.getElementsByTagName("TIME_ORDER")
for time in times:
time_slots = time.getElementsByTagName("TIME_SLOT")
for time_slot in time_slots:
ts_dict[time_slot.getAttribute("TIME_SLOT_ID")] = time_slot.getAttribute("TIME_VALUE")
time_dict_list.append(ts_dict)
#get start and end time
if start_time_list is None or end_time_list is None:
start_time_list = []
end_time_list = []
for i in range(total):
ts_dict = time_dict_list[i]
start_time_slot = None
end_time_slot = None
#go to the Shamya_looking-at-screenTier tier
tree = pr(filename[i])
annotationTree = tree.documentElement
tiers = annotationTree.getElementsByTagName("TIER")
for tier in tiers:
if tier.getAttribute("TIER_ID") == "Shamya_looking-at-screenTier":
annotations = tier.getElementsByTagName("ANNOTATION")
for annotation in annotations:
align_annotations = annotation.getElementsByTagName("ALIGNABLE_ANNOTATION")
for align_annotation in align_annotations:
if start_time_slot is None:
start_time_slot = align_annotation.getAttribute("TIME_SLOT_REF1")
end_time_slot = align_annotation.getAttribute("TIME_SLOT_REF2")
print("start_time_slot = ", start_time_slot, " end_time_slot = ", end_time_slot)
#pick start time
start_time_list.append(int(ts_dict[start_time_slot]))
end_time_list.append(int(ts_dict[end_time_slot]))
if global_mean is True:
print("Finding global mean...")
global_mean_neutral, global_mean_other = get_global_mean(facet_file, start_time_list, end_time_list)
else:
global_mean_neutral = None
global_mean_other = None
for i in range(total):
print(filename[i])
tree = pr(filename[i])
annotationTree = tree.documentElement
tiers = annotationTree.getElementsByTagName("TIER")
#put time tier to a dictionary
ts_dict = time_dict_list[i]
start_time = int(start_time_list[i])
end_time = int(end_time_list[i])
print("start_time = ", start_time, " end_time = ", end_time)
#Shamya annotations
shamya_ts = get_annotations(tiers, "Shamya_EmotionTier")
print("shamya ts ", shamya_ts)
shamya_full_ts = transform_annotation(shamya_ts, start_time, end_time,ts_dict)
print("shamya full ts ", shamya_full_ts)
#Mark annotations
mark_ts = get_annotations(tiers, "Mark_EmotionTier")
print("mark ts", mark_ts)
mark_full_ts = transform_annotation(mark_ts, start_time, end_time,ts_dict)
print("mark full ts", mark_full_ts)
#get annotation based on the block size
shamya_annot = regularize_annotations(shamya_full_ts, block, weak)
print("shamya block sized annot", shamya_annot)
mark_annot = regularize_annotations(mark_full_ts, block, weak)
print("mark block sized annot", mark_annot)
if all is True:
print(facet_file[i])
#get FACET data
#Mean Normalized
if mean is True:
if combine is True:
#has to be separate as mean of max would differ with/without the confusion column
facet = get_facet_data_combined(facet_file[i], start_time, end_time, global_mean_neutral, global_mean_other)
else:
facet = get_facet_data_all(facet_file[i], start_time, end_time, global_mean_neutral, global_mean_other)
else:
facet = get_facet_data(facet_file[i], start_time, end_time)
print("facet full", facet[:10])
if binary is True:
get_binary_emotion(facet)
facet_annot = regularize_annotations(facet, block, weak)
print("facet block sized annot", facet_annot)
#print("FACET - Neutral =", facet_annot.count('Neutral'), "Confused =", facet_annot.count('Confused'), "Other =", facet_annot.count('Other'), "Skip =", facet_annot.count('Skip'))
#print("SHAMYA - Neutral =", shamya_annot.count('Neutral'), "Confused =", shamya_annot.count('Confused'), "Other =", shamya_annot.count('Other'), "Skip =", shamya_annot.count('Skip'))
#print("MARK - Neutral =", mark_annot.count('Neutral'), "Confused =", mark_annot.count('Confused'), "Other =", mark_annot.count('Other'), "Skip =", mark_annot.count('Skip'))
#Delete skips and combine annotations
if all is True:
#change [s,m,f] if you need pairwise [m,f] or [s,f]
annot = [[s,m,f] for s,m,f in zip(shamya_annot,mark_annot, facet_annot) if s != "Skip" and m != "Skip"]
else:
annot = [[s,m] for s,m in zip(shamya_annot,mark_annot) if s != "Skip" and m != "Skip"]
#[Neutral, Confused, Other]
emot = [[0,0,0] for i in range((len(annot)))]
#update kappa matrix
emot = update_kappa_matrix(annot, emot)
emotion_kappa = emotion_kappa + emot
#if combine merge other and confused
if combine is True:
emotion_kappa = [[l[0], (l[1]+l[2])] for l in emotion_kappa]
print(len(emotion_kappa))
#compute fleiss' kappa
kappa = fleissKappa(emotion_kappa)
print(kappa)
def tweetsFromFile(filename):
tree = pr(filename)
tweets_tree = tree.documentElement
tweets = tweets_tree.getElementsByTagName("tweet")
return tweets