def configureAPI(self):
if self.title.find('.flac') != -1:
with open(os.path.join(self.path, self.title), 'rb') as audio_file:
content = audio_file.read()
self.audio = types.RecognitionAudio(content=content)
else:
self.audio = types.RecognitionAudio(uri="gs://twiml-mp3/" +
self.title + ".flac")
self.config = types.RecognitionConfig(
encoding=self.encoding,
sample_rate_hertz=self.sample_rate,
language_code=self.language_code,
enable_automatic_punctuation=self.punctuation,
enable_speaker_diarization=self.diarization,
diarization_speaker_count=self.num_speakers,
audio_channel_count=1,
use_enhanced=self.enhanced,
model=self.model,
enable_word_time_offsets=self.time_offsets,
enable_word_confidence=self.word_confidence,
max_alternatives=self.max_alternatives,
metadata=self.metadata,
speech_contexts=[types.SpeechContext(phrases=self.phrases)])
def transcribe_gcs(gcs_uri, hertz, channel):
client = speech.SpeechClient()
audio = types.RecognitionAudio(uri=gcs_uri)
config = types.RecognitionConfig(
encoding=enums.RecognitionConfig.AudioEncoding.FLAC, # flacの設定
sample_rate_hertz=int(hertz), # ヘルツは音声ファイルに合わせる
audio_channel_count=int(channel),
language_code='ja-JP', # 日本語音声の場合
enable_speaker_diarization=True, # 異なる話者の分離
enable_automatic_punctuation=True, # 句読点
speech_contexts=SELECTED_PHRASES # 音声適応ブースト
)
operation = client.long_running_recognize(config, audio)
print('Waiting for operation to complete...')
operationResult = operation.result()
filename = gcs_uri.rsplit('/', 1)[1].split('.')[0] + ".txt"
outputfilepath = os.path.join(OUTPUT_FOLDER, filename)
fout = codecs.open(outputfilepath, 'a', 'utf-8')
for result in operationResult.results:
for alternative in result.alternatives:
fout.write(u'{}\n'.format(alternative.transcript))
fout.close()
def google_transcribe_file(fp, bucket_name='prof-resp-trans'):
storage_client = google.cloud.storage.Client()
bucket = storage_client.get_bucket(bucket_name)
client = google.cloud.speech_v1p1beta1.SpeechClient()
config = types.RecognitionConfig(
encoding=enums.RecognitionConfig.AudioEncoding.ENCODING_UNSPECIFIED,
language_code='en-US',
enable_word_time_offsets=True,
enable_automatic_punctuation=True)
blob = bucket.blob(fp.name)
if not blob.exists():
new_print('Uploading File: {}'.format(fp.name))
blob.upload_from_filename(str(fp.resolve()))
new_print('Finished Uploading: {}'.format(fp.name))
else:
new_print('File already uploaded: {}'.format(fp.name))
new_print('Starting transcription...')
audio = types.RecognitionAudio(
uri='gs://{}/{}'.format(bucket_name, fp.name))
response = client.long_running_recognize(config, audio)
results = response.result()
new_print('Transciption finished')
return results
def transcribe_file_with_diarization():
audio = types.RecognitionAudio(uri=args['input'])
config = types.RecognitionConfig(
encoding=enums.RecognitionConfig.AudioEncoding.LINEAR16,
# encoding=enums.RecognitionConfig.AudioEncoding.FLAC,
sample_rate_hertz=22050,
language_code='en-US',
enable_speaker_diarization=True,
diarization_speaker_count=2,
model='phone_call')
operation = client.long_running_recognize(config, audio)
print("Waiting on response from google cloud...")
response = operation.result(timeout=720) ## 360 call 01
for result in response.results:
print("\n\n::BEGIN TRANSCRIPT::\n")
print("{}".format(result.alternatives[0].transcript))
print("\n::END TRANSCRIPT::\n\n")
print("\t\tCONFIDENCE: {} \n\n".format(
result.alternatives[0].confidence))
print("::BEGIN SPEAKER DIARIZATION::\n")
words_info = result.alternatives[0].words
for word_info in words_info:
print("{}: '{}'".format(word_info.speaker_tag, word_info.word))
print("\n::END SPEAKER DIARIZATION")
def read_audio(self, filepath):
with io.open(filepath, "rb") as f:
content = f.read()
audio = types.RecognitionAudio(content=content)
#audio = {"content": content}
return audio
def diarized_transcribe(gcred, gcs_uri, speakercount):
os.environ['GOOGLE_APPLICATION_CREDENTIALS'] = gcred
client = speech_v1p1beta1.SpeechClient()
audio = beta_types.RecognitionAudio(uri=gcs_uri)
config = beta_types.RecognitionConfig(
encoding=beta_enums.RecognitionConfig.AudioEncoding.FLAC,
language_code='en-US',
enable_speaker_diarization=True,
diarization_speaker_count=speakercount,
enable_word_time_offsets=True,
model='video',
enable_automatic_punctuation=True)
operation = client.long_running_recognize(config, audio)
response = operation.result(timeout=3600)
transcript = MessageToDict(response)
transcript = transcript.get('results')
transcript = transcript.pop()
transcript = transcript.get('alternatives')
transcript = transcript.pop()
transcript = transcript.get('words')
return transcript
def read(file_name, result_file, time=50):
# Instantiates a client
client = speech.SpeechClient()
# The name of the audio file to transcribe
file_name = os.path.join(
os.path.dirname(__file__),
file_name)
config = types.RecognitionConfig(
encoding=enums.RecognitionConfig.AudioEncoding.LINEAR16,
audio_channel_count=2,
sample_rate_hertz=44100,
enable_speaker_diarization=True,
diarization_speaker_count=2,
language_code='ko-KR')
# Loads the audio into memory
with io.open(file_name, 'rb') as audio_file:
with io.open(result_file, 'w') as f:
FILE_BYTE = 44100 * 2
read_byte = FILE_BYTE * time
while True:
content = audio_file.read(read_byte)
if not content:
break
audio = types.RecognitionAudio(content=content)
response = client.recognize(config, audio)
for result in response.results:
f.write(result.alternatives[0].transcript)
print(result.alternatives[0])
print('Transcript: {}'.format(result.alternatives[0].transcript))
def transcribe_gcs(gcs_uri):
"""Asynchronously transcribes the audio file specified by the gcs_uri."""
from google.cloud import speech
from google.cloud.speech import enums
from google.cloud.speech import types
client = speech.SpeechClient()
audio = types.RecognitionAudio(uri=gcs_uri)
config = types.RecognitionConfig(
encoding=enums.RecognitionConfig.AudioEncoding.LINEAR16,
sample_rate_hertz=8000,
language_code='en-US',
profanity_filter=True,
speech_contexts=[
speech.types.SpeechContext(
phrases=['Andy', 'Wisy', 'EEP', 'Project', 'Tom', 'Jeff'], )
],
)
operation = client.long_running_recognize(config, audio)
print('GCS -- Waiting for operation to complete...')
response = operation.result(timeout=90)
# Each result is for a consecutive portion of the audio. Iterate through
# them to get the transcripts for the entire audio file.
for result in response.results:
# The first alternative is the most likely one for this portion.
print(('Transcript: {}'.format(result.alternatives[0].transcript)))
print(('Confidence: {}'.format(result.alternatives[0].confidence)))
def transcribe_gcs(gcs_uri, hint_phrases, set_config):
"""Asynchronously transcribes the audio file specified by the gcs_uri."""
client = speech_v1p1beta1.SpeechClient()
audio = types.RecognitionAudio(uri=gcs_uri)
# hint_phrase = []
# set_config['enable_speaker_diarization'] = 'False'
print(set_config.get('enable_automatic_punctuation'))
# Set default values, check dict having each key and cast from str to each type.
config = types.RecognitionConfig(
encoding=eval(set_config.get('encoding', 'enums.RecognitionConfig.AudioEncoding.FLAC')),
sample_rate_hertz=int(set_config.get('sample_rate_hertz', 16000)),
language_code=set_config.get('language_code', 'en-US'),
enable_automatic_punctuation=eval(set_config.get('enable_automatic_punctuation', True)),
enable_speaker_diarization=eval(set_config.get('enable_speaker_diarization', False)),
diarization_speaker_count=int(set_config.get('diarization_speaker_count', 1)),
speech_contexts=[speech_v1p1beta1.types.SpeechContext(phrases=hint_phrases)])
operation = client.long_running_recognize(config, audio)
print('Waiting for operation to complete...')
response = operation.result(timeout=900)
return response
def google_transcribe(audio_file_name):
file_name = filepath + audio_file_name
frame_rate, channels = frame_rate_channel(file_name)
if channels > 1:
stereo_to_mono(file_name)
bucket_name = bucketname
source_file_name = filepath + audio_file_name
destination_blob_name = audio_file_name
upload_blob(bucket_name, source_file_name, destination_blob_name)
gcs_uri = 'gs://' + bucketname + '/' + audio_file_name
transcript = ''
client = speech.SpeechClient()
audio = types.RecognitionAudio(uri=gcs_uri)
print("Setting up configurations")
speech_context = speech.types.SpeechContext(phrases=[
"$OOV_CLASS_DIGIT_SEQUENCE", "$YEAR", "$PERCENT", "$MONEY", "$MONTH"
])
config = types.RecognitionConfig(
encoding=enums.RecognitionConfig.AudioEncoding.LINEAR16,
sample_rate_hertz=frame_rate,
language_code='en-US',
enable_speaker_diarization=True,
diarization_speaker_count=2,
speech_contexts=[speech_context],
use_enhanced=True,
model="phone_call")
# Detects speech in the audio file
print("detecting speech")
operation = client.long_running_recognize(config, audio)
response = operation.result(timeout=10000)
result = response.results[-1]
words_info = result.alternatives[0].words
tag = 1
speaker = ""
print("Assembling words")
for word_info in words_info:
if word_info.speaker_tag == tag:
speaker = speaker + " " + word_info.word
else: #Changed
transcript += "speaker {}: {}".format(tag, speaker) + '\n'
tag = word_info.speaker_tag
speaker = "" + word_info.word
transcript += "speaker {}: {}".format(tag, speaker)
delete_blob(bucket_name, destination_blob_name)
return transcript
def transcribe_file(content):
audio = types.RecognitionAudio(content=content)
response = client.recognize(config, audio)
try:
text = response.results[0].alternatives[0].transcript
return text[0].upper() + text[1:] + "?"
except IndexError:
return ""
def transcribe_gcs(gcs_uri):
from google.cloud import speech_v1p1beta1 as speech
from google.cloud.speech_v1p1beta1 import enums
from google.cloud.speech_v1p1beta1 import types
client = speech.SpeechClient()
audio = types.RecognitionAudio(uri=gcs_uri)
config = types.RecognitionConfig(
sample_rate_hertz=32000,
encoding=enums.RecognitionConfig.AudioEncoding.FLAC,
enable_speaker_diarization=True,
diarization_speaker_count=2,
language_code='ja-JP')
operation = client.long_running_recognize(config, audio)
print('Waiting for operation to complete...')
operationResult = operation.result()
d = datetime.datetime.today()
today = d.strftime("%Y%m%d-%H%M%S")
fout = codecs.open('output{}.txt'.format(today), 'a', 'shift_jis')
speaker_1_words = ""
speaker_1_s = float(0)
speaker_2_words = ""
speaker_2_s = float(0)
for word in operationResult.results[-1].alternatives[0].words:
tmp_word = u'{}'.format(word.word.split("|")[0])
start_time = float(word.start_time.seconds) + float(
word.start_time.nanos) / 1000 / 1000 / 1000
end_time = float(word.end_time.seconds) + float(
word.end_time.nanos) / 1000 / 1000 / 1000
s = end_time - start_time
if word.speaker_tag == 1:
speaker_1_s += s
speaker_1_words += tmp_word
else:
speaker_2_s += s
speaker_2_words += tmp_word
fout.write('speaker_1: \n{}\n'.format(speaker_1_words))
fout.write('s: {}\n'.format(speaker_1_s))
fout.write('speaker_2: \n{}\n'.format(speaker_2_words))
fout.write('s: {}\n'.format(speaker_2_s))
#for result in operationResult.results:
# for alternative in result.alternatives:
# fout.write(u'{}\n'.format(alternative.transcript))
fout.close()
def transcribe_file(speech_file):
"""Transcribe the given audio file asynchronously."""
from google.cloud import speech_v1p1beta1
from google.cloud.speech_v1p1beta1 import enums
from google.cloud.speech_v1p1beta1 import types
client = speech_v1p1beta1.SpeechClient()
# [START migration_async_request]
with io.open(speech_file, 'rb') as audio_file:
content = audio_file.read()
print "Using ", speech_file, ", with the below config:"
print ""
print "importing speech_v1p1beta1"
print "language_code='en-US'"
print "use_enhanced=True"
print "enable_automatic_punctuation=False"
print "enable_word_time_offsets=False"
print "profanity_filter=True"
print "sample_rate=44100hz"
print ""
print "Transcript is as follows"
audio = types.RecognitionAudio(content=content)
config = types.RecognitionConfig(
encoding=enums.RecognitionConfig.AudioEncoding.LINEAR16,
sample_rate_hertz=44100,
language_code='en-US',
use_enhanced=True,
# A model must be specified to use enhanced model.
model='phone_call',
enable_automatic_punctuation=False,
enable_word_time_offsets=False,
profanity_filter=True,
# speech_contexts=[speech.types.SpeechContext(
# phrases=['Andy', 'Wisy', 'EEP', 'Project', 'Tom', 'Jeff'],
# )],
)
# [START migration_async_response]
operation = client.long_running_recognize(config, audio)
# [END migration_async_request]
print('Waiting for operation to complete...')
response = operation.result(timeout=90)
# Each result is for a consecutive portion of the audio. Iterate through
# them to get the transcripts for the entire audio file.
for result in response.results:
# The first alternative is the most likely one for this portion.
print(u'Transcript: {}'.format(result.alternatives[0].transcript))
print('Confidence: {}'.format(result.alternatives[0].confidence))
def google_transcribe(audio_file_name):
file_name = filepath + audio_file_name
m4a_to_wav(file_name)
# The name of the audio file to transcribe
frame_rate, channels = frame_rate_channel(file_name)
if channels > 1:
stereo_to_mono(file_name)
bucket_name = bucketname
source_file_name = filepath + audio_file_name
destination_blob_name = audio_file_name
upload_blob(bucket_name, source_file_name, destination_blob_name)
gcs_uri = 'gs://' + bucketname + '/' + audio_file_name
transcript = ''
client = speech.SpeechClient()
audio = types.RecognitionAudio(uri=gcs_uri)
config = types.RecognitionConfig(
encoding=enums.RecognitionConfig.AudioEncoding.LINEAR16,
sample_rate_hertz=frame_rate,
language_code='en-IN',
enable_speaker_diarization=True,
diarization_speaker_count=2)
# Detects speech in the audio file
operation = client.long_running_recognize(config, audio)
response = operation.result(timeout=10000)
result = response.results[-1]
words_info = result.alternatives[0].words
tag = 1
speaker = ""
for word_info in words_info:
if word_info.speaker_tag == tag:
speaker = speaker + " " + word_info.word
else:
transcript += f"speaker {tag}: {speaker}\n"
tag = word_info.speaker_tag
speaker = "" + word_info.word
transcript += f"speaker {tag}: {speaker}"
delete_blob(bucket_name, destination_blob_name)
return transcript
def sample_long_running_recognize(local_file_path):
"""
Print confidence level for individual words in a transcription of a short audio
file
Separating different speakers in an audio file recording
Args:
local_file_path Path to local audio file, e.g. /path/audio.wav
"""
client = speech_v1p1beta1.SpeechClient()
# local_file_path = 'audio_files/2speaker.m4a'
# If enabled, each word in the first alternative of each result will be
# tagged with a speaker tag to identify the speaker.
enable_speaker_diarization = True
# Optional. Specifies the estimated number of speakers in the conversation.
diarization_speaker_count = 2
# The language of the supplied audio
language_code = "en-US"
config = types.RecognitionConfig(
enable_speaker_diarization=enable_speaker_diarization,
enable_automatic_punctuation=True,
#diarization_speaker_count=diarization_speaker_count,
# model='phone_call',
language_code='en-US')
with io.open(local_file_path, "rb") as f:
content = f.read()
audio = types.RecognitionAudio(content=content)
operation = client.long_running_recognize(config, audio)
print(u"Waiting for operation to complete...")
response = operation.result()
print(response.results)
for result in response.results:
# First alternative has words tagged with speakers
alternative = result.alternatives[0]
print(u"Transcript: {}".format(alternative.transcript))
# Print the speaker_tag of each word
for word in alternative.words:
print(u"Word: {}".format(word.word))
print(u"Speaker tag: {}".format(word.speaker_tag))
def transcribe_gcs(gcs_uri: str) -> Dict[str, List[Any]]:
"""Asynchronously transcribes the audio file specified by the gcs_uri."""
client = speech_v1p1beta1.SpeechClient()
# The language of the supplied audio
audio = types.RecognitionAudio(uri=gcs_uri)
operation = client.long_running_recognize(RECOGNITION_CONFIG, audio)
print(dt.now(), "Waiting for operation to complete...")
print(dt.now(), "Operation", operation.operation)
start = time()
response = operation.result(timeout=90000)
print(dt.now(), "Got response in ", time() - start)
return MessageToDict(response)
def google_transcribe(uploaded_file_path):
print("Converting: \t" + uploaded_file_path.split("/")[-1])
wav_file_path = mp3_to_wav(uploaded_file_path)
print("Converted: \t" + wav_file_path.split("/")[-1])
print("Checking frame rate: \t", wav_file_path.split("/")[-1])
frame_rate, channels = frame_rate_channel(wav_file_path)
wav_name = wav_file_path.split("/")[-1]
print("Uploading blob: \t", wav_name)
upload_blob(bucket_name, wav_file_path, wav_name)
print("Starting Transcripting: \t", wav_name)
gcs_uri = 'gs://' + bucket_name + '/' + wav_name
transcript = ''
client = speech.SpeechClient()
audio = types.RecognitionAudio(uri=gcs_uri)
config = types.RecognitionConfig(
encoding=enums.RecognitionConfig.AudioEncoding.LINEAR16,
sample_rate_hertz=frame_rate,
language_code=Language_code,
enable_speaker_diarization=True,
diarization_speaker_count=2)
# Detects speech in the audio file
operation = client.long_running_recognize(config, audio)
response = operation.result(timeout=10000)
result = response.results[-1]
words_info = result.alternatives[0].words
tag = 1
speaker = ""
for word_info in words_info:
if word_info.speaker_tag == tag:
speaker = speaker + " " + word_info.word
else:
transcript += "speaker {}: {}".format(tag, speaker) + '\n'
tag = word_info.speaker_tag
speaker = "" + word_info.word
transcript += "speaker {}: {}".format(tag, speaker)
print("Deleting blob: \t", wav_name)
delete_blob(bucket_name, wav_name)
return transcript
def transcribe_interviews(gcs_uri):
"""Asynchronously transcribes the audio file specified by the gcs_uri."""
from google.cloud import speech_v1p1beta1 as speech
from google.cloud.speech_v1p1beta1 import enums
from google.cloud.speech_v1p1beta1 import types
client = speech.SpeechClient()
audio = types.RecognitionAudio(uri=gcs_uri)
config = types.RecognitionConfig(
encoding=enums.RecognitionConfig.AudioEncoding.
LINEAR16, #this is for WAV files, you can use multiple types
sample_rate_hertz=44100,
language_code='en-US',
#use_enhanced=True, #can only use if your ethics allows sending data offsite
enable_word_time_offsets=True,
enable_automatic_punctuation=True,
model='Video') # change this if you're doing focus groups
operation = client.long_running_recognize(config, audio)
print('A little man is now listening and transcribing...')
response = operation.result(timeout=90000000)
f = open("Interview 1.txt",
"w") #can change the file name that the text gets written to
for i, result in enumerate(response.results):
alternative = result.alternatives[0]
f.write(str(i) + '\n')
f.write('{}'.format(
str(
time.strftime(
'%H:%M:%S',
time.gmtime(int(
alternative.words[0].start_time.seconds))))))
f.write(' --> ')
f.write('{}'.format(
str(
time.strftime(
'%H:%M:%S',
time.gmtime(int(alternative.words[-1].end_time.seconds)))))
+ '\n')
#f.write('speaker {} :'.format(alternative.words[0].speaker_tag))
f.write(u'{}'.format(alternative.transcript) + '\n\n')
def transcribe_gcs(gcs_uri, encoding="LINEAR16", sample_rate=16000):
"""Asynchronously transcribes the audio file specified by the gcs_uri."""
client = speech.SpeechClient()
audio = types.RecognitionAudio(uri=gcs_uri)
config = types.RecognitionConfig(
encoding=ENCODINGS[encoding],
sample_rate_hertz=sample_rate,
language_code='ko-KR',
enable_automatic_punctuation=True,
enable_word_time_offsets=True,
enable_speaker_diarization=True, # 한국어 지원 안됨 (speaker_tag가 모두 동일인으로 분류됨)
diarization_speaker_count=3)
operation = client.long_running_recognize(config, audio)
print('Waiting for operation to complete...')
response = operation.result(timeout=300)
words_with_tags = []
transcripts = []
# Each result is for a consecutive portion of the audio.
# Iterate through them to get the transcripts for the entire audio file.
for result in response.results:
# The first alternative is the most likely one for this portion.
print(u'Transcript: {}'.format(result.alternatives[0].transcript))
transcripts.append(result.alternatives[0].transcript
) # punctuation 포함된 문장을 사용하기 위해 저장
for words in result.alternatives[0].words:
word = words.word
start_time = round(
words.start_time.seconds + words.start_time.nanos * 1e-9, 3)
end_time = round(
words.end_time.seconds + words.end_time.nanos * 1e-9, 3)
speaker_tag = words.speaker_tag
words_with_tags.append(
[word, start_time, end_time,
speaker_tag]) # [word, start_time, end_time, speaker_tag]
print()
return words_with_tags, transcripts
def run_speaker_diarization(self, audio_uri, audio_ch, audio_sr,
max_speakers):
logger.info('Performing Speaker Diarization for {}'.format(audio_uri))
drzr_config = types.SpeakerDiarizationConfig(
enable_speaker_diarization=True, max_speaker_count=max_speakers)
config = speech.types.RecognitionConfig(
language_code="en-US",
sample_rate_hertz=int(audio_sr),
encoding=enums.RecognitionConfig.AudioEncoding.MP3,
audio_channel_count=int(audio_ch),
enable_word_time_offsets=True,
model="video",
enable_automatic_punctuation=False,
diarization_config=drzr_config)
audio_file = types.RecognitionAudio(uri=audio_uri)
operation = self.client.long_running_recognize(config=config,
audio=audio_file)
res = operation.result()
return res
def long_transcribe_gcs(self, gcs_uri):
print("Transcribing... (This may take a while)")
audio = types.RecognitionAudio(uri=gcs_uri)
operation = self.config.client.long_running_recognize(self.config.recognition_config, audio)
response = operation.result(timeout=90)
# Each result is for a consecutive portion of the audio. Iterate through
# them to get the transcripts for the entire audio file.
text = ""
for result in response.results:
speaker_num = result.alternatives[0].words[0].speaker_tag
speaker = self.config.speakers[speaker_num - 1]
text += speaker + result.alternatives[0].transcript + '\n'
# The first alternative is the most likely one for this portion.
# print(u'Transcript: {}'.format(result.alternatives[0].transcript))
# print('Confidence: {}'.format(result.alternatives[0].confidence))
return text
def speech_to_text(gcs_uri):
#ltt_context = open('context.txt', 'r').read().split('\n')
client = speech_v1p1beta1.SpeechClient()
audio = types.RecognitionAudio(uri=gcs_uri)
#speech_contexts_element = {"phrases": ltt_context, "boost": 11}
#speech_contexts = [speech_contexts_element]
config = {
"encoding": enums.RecognitionConfig.AudioEncoding.MP3,
"sample_rate_hertz": 48000,
"language_code": 'en-US',
#"speech_contexts": speech_contexts,
"max_alternatives": 11,
"model": "video",
"enable_word_confidence": True,
"enable_word_time_offsets": True,
"enable_automatic_punctuation": True
}
operation = client.long_running_recognize(config, audio)
print('Speech-to-Text running.')
response = operation.result()
return (response)
def transcribe_audio():
I = 4
try:
alter_db(I)
except Exception as e:
print e
conn, c = get_db(databases[I])
conn.execute("SELECT %s FROM %s LIMIT 1 " %
(database_keys[I], database_names[I]))
rows = conn.fetchall()
for row in rows:
#print row
audio_file = "%s%s/%s.wav" % (DIR, FILE_NAMES[I], row[0])
audio_file = copy_files(audio_file, row[0])
print audio_file
if os.path.isfile(audio_file):
with io.open(audio_file, 'rb') as audio_file2:
content = audio_file2.read()
audio_data = types.RecognitionAudio(content=content)
# try:
response = client.recognize(config, audio_data)
transcript = "-1"
confidence = 0
if len(response.results):
result = response.results[0]
if result.alternatives:
transcript = result.alternatives[0].transcript
confidence = result.alternatives[0].confidence
print response
#score = get_jaccard_sim()
sql = "UPDATE " + database_names[
I] + " SET transcript=?,confidence=? WHERE Filename=? "
print sql
conn.execute(sql, (transcript, confidence, row[0]))
c.commit()
def wordTimeOffsets(filename, phrases, flag):
client = speech.SpeechClient()
if flag == 0:
speech_file = "/home/malkaiv/project/final/recordings/wav/" + str(filename) + ".wav"
else:
print("in flag = 1")
speech_file = "/home/malkaiv/project/final/recordings/wordsWav/" + str(filename) +".wav"
with io.open(speech_file, 'rb') as audio_file:
content = audio_file.read()
audio = types.RecognitionAudio(content=content)
with wave.open(speech_file, 'rb') as wave_file:
frame_rate = wave_file.getframerate()
# speechText = "please call Stella ask her to bring these things with her from the store \
# six spoons of fresh snow peas five thick slabs of blue cheese and maybe a snack for her \
# brother Bob we also need a small plastic snake and a big toy frog for the kids she can \
# scoop these things into three red bags and we will go meet her Wednesday at the train station"
# phrases = speechText.lower().split()
boost = 20.0
speech_contexts_element = {"phrases": phrases, "boost": boost}
speech_contexts = [speech_contexts_element]
config = types.RecognitionConfig(
speech_contexts=speech_contexts,
encoding=enums.RecognitionConfig.AudioEncoding.LINEAR16,
sample_rate_hertz=frame_rate,
language_code="en-US",
enable_word_time_offsets=True,
enable_word_confidence=True,
)
# print(client.recognize(config=config, audio=audio))
# first = response.results[0].alternatives[0]
# print("{}\n{}".format(first.transcript, first.confidence))
return client.recognize(config=config, audio=audio)
def transcribe(data):
idx, file = data
num, _ = file.split('.')
text_script = ""
name = file
print(file + " - started")
# Loads the audio into memory
with io.open(fileFullPathname+'\\'+file, 'rb') as audio_file:
content = audio_file.read()
audio = types.RecognitionAudio(content=content)
# Transcribe audio file
# Detects speech in the audio file
client = speech.SpeechClient()
response = client.recognize(config, audio)
for result in response.results:
text_script += result.alternatives[0].transcript
print(name + " - done")
return {
"idx": num,
"text": text_script
}
def transcribe_gcs(gcs_uri, phrase_hints=[], language_code="en-US"):
"""Asynchronously transcribes the audio file specified by the gcs_uri."""
client = speech.SpeechClient()
phrases = phrase_hints
audio = types.RecognitionAudio(uri=gcs_uri)
config = types.RecognitionConfig(
encoding=enums.RecognitionConfig.AudioEncoding.LINEAR16,
sample_rate_hertz=16000,
language_code=language_code,
enable_word_time_offsets=True,
model = 'video',
diarization_speaker_count=2,
enable_automatic_punctuation=True,
use_enhanced=True,
enable_speaker_diarization=True,
speech_contexts=[speech.types.SpeechContext(phrases=phrases)]
)
operation = client.long_running_recognize(config, audio)
transcription_response = operation.result(timeout=90000)
return transcription_response.results
def transcribe_streaming_from_file(speech_file):
"""Transcribe the given audio file."""
from google.cloud import speech_v1p1beta1 as speech
from google.cloud.speech_v1p1beta1 import enums
from google.cloud.speech_v1p1beta1 import types
client = speech.SpeechClient()
with io.open(speech_file, 'rb') as audio_file:
content = audio_file.read()
audio = types.RecognitionAudio(content=content)
config = types.RecognitionConfig(
encoding=enums.RecognitionConfig.AudioEncoding.LINEAR16,
language_code='en-US')
response = client.recognize(config, audio)
# Each result is for a consecutive portion of the audio. Iterate through
# them to get the transcripts for the entire audio file.
for result in response.results:
# The first alternative is the most likely one for this portion.
# The first alternative is the most likely one for this portion.
return result.alternatives[0].transcript
def translate_with_timestamps(self, gs_uri, encoding, mode, hint):
audio = types.RecognitionAudio(uri=gs_uri)
config = types.RecognitionConfig(
encoding=encoding,
language_code=mode,
enable_word_time_offsets=True,
speech_contexts=[types.SpeechContext(phrases=hint)],
enable_word_confidence=True)
operation = self.client.long_running_recognize(config=config,
audio=audio)
results = []
for result in operation.result().results:
alternatives = result.alternatives
if not alternatives:
continue
alternative = alternatives[0]
results.append([alternative.transcript, alternative.confidence])
for word_info in alternative.words:
word = word_info.word
start_time = word_info.start_time.seconds + word_info.start_time.nanos * 1e-9
end_time = word_info.end_time.seconds + word_info.end_time.nanos * 1e-9
confidence = word_info.confidence
results.append([word, start_time, end_time, confidence])
return results
def transcribe_file(speech_file):
"""Transcribe the given audio file asynchronously."""
from google.cloud import speech_v1p1beta1
from google.cloud.speech_v1p1beta1 import enums
from google.cloud.speech_v1p1beta1 import types
client = speech_v1p1beta1.SpeechClient()
# [START migration_async_request]
with io.open(speech_file, 'rb') as audio_file:
content = audio_file.read()
print('Using ', speech_file, ', with the below config:')
print("")
print("importing speech_v1p1beta1")
print("language_code='en-US'")
print("use_enhanced=True")
print("enable_automatic_punctuation=False")
print("enable_word_time_offsets=False")
print("profanity_filter=True")
print("sample_rate=44100hz")
print("")
print("Transcript is as follows")
print("Trans_Output_" + foldernametime)
audio = types.RecognitionAudio(content=content)
config = types.RecognitionConfig(
encoding=enums.RecognitionConfig.AudioEncoding.LINEAR16,
sample_rate_hertz=44100,
language_code='en-US',
use_enhanced=True,
# A model must be specified to use enhanced model.
model='phone_call',
enable_automatic_punctuation=False,
enable_word_time_offsets=False,
profanity_filter=True,
# speech_contexts=[speech.types.SpeechContext(
# phrases=['Andy', 'Wisy', 'EEP', 'Project', 'Tom', 'Jeff'],
# )],
)
# [START migration_async_response]
operation = client.long_running_recognize(config, audio)
# [END migration_async_request]
os.chdir("Trans_Output_" + foldernametime)
with open("output_transcription.txt", "a") as myfile:
#print('File -- before write file')
#myfile.write(datetime.now().strftime("%Y-%m-%d %H:%M:%S")+ "\n")
print('File -- Waiting for operation to complete...')
response = operation.result(timeout=90)
# Each result is for a consecutive portion of the audio. Iterate through
# them to get the transcripts for the entire audio file.
for result in response.results:
# The first alternative is the most likely one for this portion.
print(('Transcript: {}'.format(result.alternatives[0].transcript)))
print(('Confidence: {}'.format(result.alternatives[0].confidence)))
with open("output_transcription.txt", "a") as myfile:
myfile.write(datetime.now().strftime("%Y-%m-%d %H:%M:%S -"))
myfile.write(
(' {}'.format(result.alternatives[0].transcript)) + "\n")
#myfile.write(('Confidence: {}'.format(result.alternatives[0].confidence))+ "\n")
with open("output_transcription.txt", "a") as myfile:
myfile.write('')
myfile.close()
# [END migration_async_response]
exit()
async def speech_to_text(queue):
"""
This function implements the translation from speech to text with online and offline services, and compute the
emotion related to the speech
:param queue: process shared queue
"""
kb_client = kb.KnowledgeBaseClient(False)
kb_ID = (kb_client.register())['details']
kb_client.registerTags(kb_ID, { 'AV_IN_TRANSC_EMOTION' : {'desc' : 'text from audio', 'doc' : """```json\n{\n\t"tag": 'AV_IN_TRANSC_EMOTION',\n\t"timestamp": int,\n\t"ID": int,\n\t"text": string,\n\t"language": string,\n\t"valence": float,\n\t"arousal": float\n}```"""} })
# Create new recogniers for all the services used
r = sr.Recognizer()
google_client = None
try:
google_client = speech.SpeechClient()
except exceptions.DefaultCredentialsError as e:
log.error("Failed to authenticate with Google Cloud Speech recognition" + str(e))
except :
log.error("Unexpected error. Failed to authenticate with Google Cloud Speech recognition:"+ str(sys.exc_info()[0]))
with ThreadPoolExecutor() as executor:
while True:
# Data stored in the queue contain all the information needed to create AudioData object
timestamp, channels, sampleRate, bitPerSample, data = await queue.get()
audio = sr.AudioData(data, sampleRate, bitPerSample/8)
audio_gc = types.RecognitionAudio(content=data)
# Compute the transcription of the audio
google_cloud = executor.submit(recognize, "google-cloud", audio_gc, google_client)
google = executor.submit(recognize, "google", audio, r)
sphinx = executor.submit(recognize, "sphinx", audio, r)
# Compute the emotion related to the audio
#emotion = executor.submit(sentimental_analizer.emotion_from_speech, sampleRate, audio, log)
res = google_cloud.result()
if res["error"] is None:
# Add to KB Google cloud speech recognition result with timestamp and ID
log.info("Insert into KB --> Google cloud speech recognition result: " + str(res["text"]))
else:
log.error("Google cloud speech recognition retrieved an error: " + str(res["error"]))
res = google.result()
if res["error"] is None:
# Add to KB Google result with timestamp and ID
log.info("Insert into KB --> Google result: " + str(res["text"]))
else:
log.error("Google retrieved an error: " + str(res["error"]))
res = sphinx.result()
if res["error"] is None:
# Add to KB Sphinx result with timestamp and ID
log.info("Insert into KB --> Sphinx result: " + str(res["text"]))
else:
log.error("Sphinx retrieved an error: " + str(res["error"]))
emotion = {"valence": 1, "arousal":1} #emotion.result()
myID = 'stt'
if res["error"] is None:
# Add to KB that the transcription of the audio
kb_client.addFact(kb_ID, 'AV_IN_TRANSC_EMOTION', 1, 100, {"tag": 'AV_IN_TRANSC_EMOTION',
"timestamp": timestamp,
"ID": timestamp,
"text": res["text"],
"language": res["lang"],
"valence": emotion["valence"],
"arousal": emotion["arousal"]
})
# TODO adjust "text_f_audio", 2, 50, 'false'
else:
# Add to KB that none of google and sphinx retrieved a result
log.critical("Insert into KB that no Google or Sphinx result")
kb_client.addFact(kb_ID, 'AV_IN_TRANSC_EMOTION', 1, 100, {"tag": 'AV_IN_TRANSC_EMOTION',
"timestamp": timestamp,
"ID": timestamp,
"text": "",
"language": res["lang"],
"valence": emotion["valence"],
"arousal": emotion["arousal"]
})
