2016-09-26 70 views
0

我们正在使用的API无会话和蟒蛇,我们已经把“继续:真”:语音到文本沃森中断的沉默

高清make_completed_audio_request(URL,API_NAME =无,语言=无,时间像这样的参数=无):

username, password, endpoint, lan=select_voice_api(name=API_name, language=language) 
audio = get_complete_audio(url, api_name=API_name, time=time) 
endpoint=get_post_endpoint(url=endpoint, api_name=API_name) 
if audio: 
    list_audio=get_speakers_by_audio(audio[1].name) 
    headers={'content-type': audio[2]} 
    params = {'model': lan, 
     'continuous':True, 
       'timestamps': True} 
    if language and (API_name == 'watson' or API_name == 'WATSON'): 
     print 'enviando request' 
     response = requests.post(url=endpoint, auth=(username, password), 
      params=params, data=audio[1], headers=headers) 
     print 'cladificando error' 
     error_clasifier(code=response.status_code) 
    else: 
     response = requests.post(url=endpoint, auth=(username, password), 
      params=params, data=audio[1], headers=headers) 
     error_clasifier(code=response.status_code) 
    if response: 
    return response, list_audio, True, None 
else: 
    return None, None, False, None 

但它仍然无法正常工作,它减少转录在它创立

什么我做错了第一沉默?有没有另一种方式将其发送到API?

+0

还有很多其他API更准确,可识别长文件。 –

+0

感谢评论@NikolayShmyrev,但我们决定使用watson作为kewywords功能。 – avi1074

回答

0

我正在使用watson_developer_cloud API。它易于使用,更重要的是 - 它的工作原理。这里是代码示例:

import json 

from os.path import join, dirname 

from watson_developer_cloud import SpeechToTextV1 

speech_to_text = SpeechToTextV1(
    username="yourusername", 
    password="yourpassword", 
    x_watson_learning_opt_out=False) 

with open(join(dirname(__file__), 'test.wav'), 'rb') as audio_file: 

    data = json.dumps(speech_to_text.recognize(audio_file, content_type='audio/wav', word_confidence=True, continuous=True, word_alternatives_threshold=0, max_alternatives=10)) 
+0

它工作完美!谢谢 – avi1074

相关问题