def do_IS_speakers():
output = []
Productions = ['2','3']
for w in Words:
mod_path = os.path.join(MODEL_DIR,'modeltalker_%s.wav' %w)
mod_env = calc_envelope(mod_path,**opt_dict)
print(w)
for s in Speakers:
if 'IS' not in s:
continue
if s == 'modeltalker':
continue
sp = s.split('_')[0]
s_dir = os.path.join(JAM_DIR,s)
for p in Productions:
base_path = os.path.join(s_dir,'%s_%s1is.wav' % (sp,w))
shad_path = os.path.join(s_dir,'%s_%s%sis.wav' % (sp,w,p))
if not os.path.isfile(base_path):
continue
if not os.path.isfile(shad_path):
continue
base_env = calc_envelope(base_path,**opt_dict)
shad_env = calc_envelope(shad_path,**opt_dict)
b_to_m_sim = envelope_match(mod_env,base_env)
s_to_m_sim = envelope_match(mod_env,shad_env)
output.append([sp,p,w,b_to_m_sim,s_to_m_sim])
with open(os.path.join(BASE_DIR,'jam_output8IS.txt'),'w') as f:
csvw = csv.writer(f,delimiter='\t')
csvw.writerow(['Shadower_number','Block',
'Word','base_to_mod_env_sim','shad_to_mod_env_sim',
#'mfcc_shad_to_mod','mfcc_mod_to_shad',
#'spec_shad_to_mod','spec_mod_to_shad',
#'mfcc_base_to_mod','mfcc_mod_to_base',
#'spec_base_to_mod','spec_mod_to_base'
])
for l in output:
csvw.writerow(l)
def convert_tokens_to_graph(s):
print('%s' % (s))
filepath = os.path.join(fetch_buckeye_resource('Graphs'),'%s.txt' % (str(s),))
if os.path.isfile(filepath):
return None
g = nx.Graph()
tokens = WordToken.objects.select_related('WordType','Dialog__Speaker','Category').filter(Dialog__Speaker = s)
#tokens = tokens.exclude(Category__CategoryType__in = ['Pause','Disfluency','Other'])
tokens = tokens.filter(WordType__Label__in = GOOD_WORDS)
print(len(tokens))
if len(tokens) == 0:
return None
g.add_nodes_from([(x.pk,{'Word':x.WordType.Label,
'Frequency':x.WordType.get_frequency(),
'ND': x.WordType.get_ND(),
'DialogPlace':x.get_dialog_place(),
'Duration':x.get_duration(),
'Speaker':str(s),
'SpeakerGender':s.Gender,
'SpeakerAge': s.Age,
'PrevSpeakRate' :x.get_previous_speaking_rate(),
'FollSpeakRate' : x.get_following_speaking_rate(),
'AvgSpeakRate' : s.get_avg_speaking_rate(),
'PrevCondProb' : x.get_previous_cond_prob(),
'FollCondProb' : x.get_following_cond_prob(),
'Repetitions' : x.get_repetitions(),
'wasRepeatedRecently' : x.get_recent_repetition(),
'OrthoLength' : len(x.WordType.Label),
'PhonoLength': x.WordType.UR.count(),
}) for x in tokens if x.is_acceptable()])
for i in range(len(tokens)):
envone = tokens[i].get_envelope()
for j in range(i+1,len(tokens)):
envtwo = tokens[j].get_envelope()
sim = envelope_match(envone,envtwo)
g.add_edge(tokens[i].pk,tokens[j].pk,weight = sim)
path = fetch_temp_resource('buckeye-wt-%d.wav' % tokens[i].pk)
try:
os.remove(path)
except FileNotFoundError:
pass
pickle.dump(g,open(filepath,'wb'))