-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathlisten_and_punctuate.py
155 lines (139 loc) · 6.8 KB
/
listen_and_punctuate.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
import microphone_recognition as mr
import speech_recognition as sr
import sys
sys.path.insert(0, "punkProse")
import models
import punctuator
import utilities
import theano
import os
import yaml
import copy
sys.path.insert(0, "Proscript/proscript")
from proscript.proscript import Word, Segment, Proscript
from proscript.utilities import utils
model_file_w = "models/punkProse_model_eng_wordonly.pcl"
model_file_wPOSpmf = "models/punkProse_model_eng_prosodic.pcl"
config_file = "punkProse/parameters.yaml"
working_dir_name = mr.WORKING_DIR
if __name__ == '__main__':
print("Loading configurations file...")
with open(config_file, 'r') as ymlfile:
config = yaml.load(ymlfile)
print("Loading model parameters...")
net_wPOSpmf, inputs_wPOSpmf, input_feature_names_wPOSpmf, _ = models.load(model_file_wPOSpmf, 1)
net_w, inputs_w, input_feature_names_w, _ = models.load(model_file_w, 1)
#net_wp, inputs_wp, input_feature_names_wp, _ = models.load(model_file_wp, 1)
print("Building model...")
predict_wPOSpmf = theano.function(inputs=inputs_wPOSpmf, outputs=net_wPOSpmf.y)
predict_w = theano.function(inputs=inputs_w, outputs=net_w.y)
#predict_wp = theano.function(inputs=inputs_wp, outputs=net_wp.y)
print("Loading vocabularies...")
#vocabulary_dict, leveler_dict = punctuator.load_dictionaries(config, input_feature_names_w)
vocabulary_dict, leveler_dict = punctuator.load_dictionaries(config, input_feature_names_wPOSpmf)
#load recognition tools
r = sr.Recognizer()
while True:
key_input = input("Press R to record, O to open audio file, P to open proscript file, Q to quit...")
print(key_input)
if key_input == "q" or key_input=="Q":
sys.exit()
elif key_input == "r" or key_input=="R" or key_input=="":
#run recognition
proscript_path, transcription = mr.run_microphone_recognizer(working_dir_name, r)
elif key_input == "o" or key_input=="O":
file_input = input("Input filename: ")
file_input = file_input.strip()
if os.path.isfile(file_input) and file_input.endswith(".wav"):
proscript_path, transcription = mr.run_microphone_recognizer(working_dir_name, r, wav_in = file_input.strip())
else:
print("File doesn't exist")
proscript_path = None
elif key_input == "p" or key_input=="P":
csv_file_input = input("Input filename: ")
csv_file_input = csv_file_input.strip()
if os.path.isfile(csv_file_input) and csv_file_input.endswith(".csv"):
wav_file_input = os.path.join(os.path.dirname(csv_file_input), os.path.splitext(os.path.basename(csv_file_input))[0] + '.wav')
print(wav_file_input)
if os.path.isfile(wav_file_input):
proscript_path, transcription = mr.run_microphone_recognizer(working_dir_name, r, wav_in=wav_file_input, csv_in=csv_file_input) #only used for copying the audio to working dir
transcription = None
else:
print("Proscript is not accompanied with its audio")
else:
print("File doesn't exist")
proscript_path = None
else:
proscript_path = None
#punctuate the input
if proscript_path:
#punctuate proscript
proscript_data = utilities.read_proscript(proscript_path, add_end=True)
if transcription == None:
transcription = ' '.join(proscript_data['word'])
punctuated_proscript_data_w = copy.copy(proscript_data)
punctuated_transcript_w = punctuator.restore_unsequenced_test_data( punctuated_proscript_data_w,
vocabulary_dict=vocabulary_dict,
leveler_dict=leveler_dict,
predict_function=predict_w,
input_feature_names=input_feature_names_w,
sequence_length=config["SAMPLE_SIZE"],
readable_format=True)
# punctuated_proscript_data_wp = copy.copy(proscript_data)
# punctuated_transcript_wp = punctuator.restore_unsequenced_test_data( punctuated_proscript_data_wp,
# vocabulary_dict=vocabulary_dict,
# leveler_dict=leveler_dict,
# predict_function=predict_wp,
# input_feature_names=input_feature_names_w,
# sequence_length=config["SAMPLE_SIZE"],
# readable_format=True)
punctuated_proscript_data_wPOSpmf = copy.copy(proscript_data)
punctuated_transcript_wPOSpmf = punctuator.restore_unsequenced_test_data( punctuated_proscript_data_wPOSpmf,
vocabulary_dict=vocabulary_dict,
leveler_dict=leveler_dict,
predict_function=predict_wPOSpmf,
input_feature_names=input_feature_names_wPOSpmf,
sequence_length=config["SAMPLE_SIZE"],
readable_format=True)
#Write returned proscripts to file in the input dir.
#working_dir = os.path.dirname(proscript_path)
recording_id = os.path.basename(proscript_path).split('.')[0]
#print out stuff
print("################################################################")
print("################# UNPUNCTUATED TRANSCRIPTION ###################")
print(transcription)
unpunctuated_proscript_id = recording_id + '.0'
unpunctuated_proscript_file = os.path.join(working_dir_name, unpunctuated_proscript_id + '.csv')
proscript_data['punctuation_after'] = [''] * len(proscript_data['punctuation_after'])
p0 = Proscript()
p0.from_dict(proscript_data, unpunctuated_proscript_id)
p0.to_csv(unpunctuated_proscript_file)
print("################# PUNCTUATED WITH punkProse ####################")
print("Model 1: word")
print(punctuated_transcript_w)
punctuated_proscript_id = recording_id + '.1'
punctuated_proscript_file = os.path.join(working_dir_name, punctuated_proscript_id + '.csv')
p1 = Proscript()
p1.from_dict(punctuated_proscript_data_w, punctuated_proscript_id)
p1.to_csv(punctuated_proscript_file)
print("----------------------------------------------------------------")
print("----------------------------------------------------------------")
# print("Model 2: word+pause")
# print(punctuated_transcript_wp)
# punctuated_proscript_id = recording_id + '.2'
# punctuated_proscript_file = os.path.join(working_dir_name, punctuated_proscript_id + '.csv')
# p2 = Proscript()
# p2.from_dict(punctuated_proscript_data_wp, punctuated_proscript_id)
# p2.to_csv(punctuated_proscript_file)
# print("----------------------------------------------------------------")
# print("----------------------------------------------------------------")
print("Model 2: word + POS + pause + f0_mean")
print(punctuated_transcript_wPOSpmf)
print("################################################################")
punctuated_proscript_id = recording_id + '.2'
punctuated_proscript_file = os.path.join(working_dir_name, punctuated_proscript_id + '.csv')
p3 = Proscript()
p3.from_dict(punctuated_proscript_data_wPOSpmf, punctuated_proscript_id)
p3.to_csv(punctuated_proscript_file)
else:
print("Can't recognize")