Google Speech API 转录电子邮件答案

【问题标题】：Google Speech API transcribe emailGoogle Speech API 转录电子邮件
【发布时间】：2020-10-06 14:22:32
【问题描述】：

我在使用 Google Speech REST API 转录电子邮件时遇到问题。我能得到的最好的是大多数电子邮件地址，但是 Google Speech 忽略了“dot”和“dot com”。例如 first.last@gmail.com 返回“gmail 的 First Last”。如果我说“句号”而不是“点”，我至少会得到“第一个。最后在 gmail”。我正在使用以下内容：

{
  "config": {
      "encoding": "MULAW",
      "sampleRateHertz": 8000,
      "languageCode": "en-US",
      "maxAlternatives": 0,
      "profanityFilter": true,
      "enableWordTimeOffsets": false,
      "model": "phone_call",
      "useEnhanced": true
  },
  "audio": {
      "content":"&&NameBase64&&"
  }
}

我尝试添加“点”作为语音上下文，没有任何变化。 “.”、“.com”、“com”和“kom”也没有改变结果。

{
  "config": {
      "encoding": "MULAW",
      "sampleRateHertz": 8000,
      "languageCode": "en-US",
      "maxAlternatives": 1,
      "profanityFilter": true,
      "enableWordTimeOffsets": false,
      "model": "phone_call",
      "useEnhanced": true,
      "speechContexts": [{
        "phrases": ["dot"],
        }],
  },
  "audio": {
      "content":"Base64Recording"
  }
}

我尝试添加字母数字语音上下文并将其拼写出来，但结果非常糟糕。

关于如何获得“。”的任何想法或“dot”和“com”出现在转录中将不胜感激。

【问题讨论】：

标签： speech-to-text google-speech-api google-cloud-speech

【解决方案1】：

您是否尝试过为该短语提供提升值？我面临同样的问题，我注意到增加提升值有助于识别“点”这个词。提升值通常在 0 到 20 之间，但应用任何高于 10 的值都有助于识别“点”。

这是一个例子：-

  "config": {
      "encoding": "MULAW",
      "sampleRateHertz": 8000,
      "languageCode": "en-US",
      "maxAlternatives": 1,
      "profanityFilter": true,
      "enableWordTimeOffsets": false,
      "model": "phone_call",
      "useEnhanced": true,
      "speechContexts": [{
        "phrases": ["dot"],
        "boots": 15.0
        }],
  },
  "audio": {
      "content":"Base64Recording"
  }
}

您还可以在上下文中有多个键值对，每个键值对都有不同的提升值。例如，这是我用来检测电子邮件地址的：-

[{
    phrases: ["$OOV_CLASS_ALPHANUMERIC_SEQUENCE"],
    boost: 14.0
},
{
    phrases: ["gmail.com","yahoo.com","aol.com","outlook.com"],
    boost:5.0
},
{
    phrases: ["com",".","c o m",".com","dotcom","dot com","dot","at","at the rate","@"],
    boost: 10.0
},
{
    phrases: ["org","io","dot org","dot io","gov","dot gov","net","dot net","co","dot co"],
    boost:8.0
},
{
    phrases: ["$OOV_CLASS_DIGIT_SEQUENCE","8","naught","z","zed","zee","zz","d","aa","ae","ee","oo","ii","ay","eh","ahh","ah","ze","dee",
"1","2","3","4","5","6","7","8","9","0","zero",],
    boost: -20.0
}
]

请注意，具有负提升值的短语将有助于清除经常被误解的单词。

【讨论】：