Script que obtiene la transcripción de un audio alojado en Cloud Storage y busca coincidencias con expresiones regulares

audio-transcription-regex.py

import re
import audioread
from google.cloud import speech
from google.cloud.speech import enums
from google.cloud.speech import types

AUDIO_CHANNELS = 2
AUDIO_RATE = 16000
BLOB_PATH = "audio.flac"
BUCKET_NAME = "audio-devlife"
LANG_CODE = 'es-MX'

ORIGINAL_PHRASE = r'(?P<phrase>(vi|bi){1}da de programador)'
POSSIBLE_PHRASE = r'(?P<phrase>(vi|bi){1}.{0,3}(de)? programador)'
DEV_PHRASE = r'programador'


with audioread.audio_open(BLOB_PATH) as f:
    AUDIO_CHANNELS = int(f.channels)
    AUDIO_RATE = int(f.samplerate)

gcs_uri = "gs://{}/{}".format(BUCKET_NAME, BLOB_PATH)
client = speech.SpeechClient()

audio = types.RecognitionAudio(uri=gcs_uri)

config = types.RecognitionConfig(
    encoding=enums.RecognitionConfig.AudioEncoding.FLAC,
    language_code=LANG_CODE,
    audio_channel_count=AUDIO_CHANNELS,
    enable_separate_recognition_per_channel=True)

operation = client.long_running_recognize(config, audio)
response = operation.result()
contador_phx = 0
contador_php = 0
contador_dev = 0

for result in response.results:
    for alternative in result.alternatives:
        print('=' * 20)
        text_block = alternative.transcript
        print(text_block)
        phx = re.findall(ORIGINAL_PHRASE, text_block,
                         re.MULTILINE & re.IGNORECASE)
        php = re.findall(POSSIBLE_PHRASE, text_block,
                         re.MULTILINE & re.IGNORECASE)
        devp = re.findall(DEV_PHRASE, text_block,
                          re.MULTILINE & re.IGNORECASE)

        if phx:
            contador_phx += len(phx)

        if php:
            contador_php += len(php)

        if devp:
            contador_dev += len(devp)

print("\nFRASE ORIGINAL: {}\tFRASE POSIBLE: {}\t PALABRA PROGRAMADOR:{}".
      format(contador_phx, contador_php, contador_dev))