启用字词级置信度

您可以指定 Speech-to-Text 为转录文本中的每个字词指示准确度(置信度)值。

字词级置信度

Speech-to-Text 在转录音频剪辑的同时,也会测量响应的准确度。Speech-to-Text 发送的响应以介于 0.0 与 1.0 之间的数值表示整个转录请求的置信度。 以下代码示例展示了 Speech-to-Text 返回的一个示例置信度值。

{
  "results": [
    {
      "alternatives": [
        {
          "transcript": "how old is the Brooklyn Bridge",
          "confidence": 0.96748614
        }
      ]
    }
  ]
}

除了整个转录的置信度之外,Speech-to-Text 还可以提供转录内各个字词的置信度。在这种情况下,响应会在转录中包含 WordInfo 详细信息,表明各个字词的置信度,如以下示例所示。

{
  "results": [
    {
      "alternatives": [
        {
          "transcript": "how old is the Brooklyn Bridge",
          "confidence": 0.98360395,
          "words": [
            {
              "startTime": "0s",
              "endTime": "0.300s",
              "word": "how",
              "confidence": SOME NUMBER
            },
            ...
          ]
        }
      ]
    }
  ]
}

在请求中启用字词级置信度

以下代码段演示了如何使用本地和远程文件在发送给 Speech-to-Text 的转录请求中启用字词级置信度功能。

使用本地文件

协议

如需了解完整的详细信息,请参阅 speech:recognize API 端点。

如需执行同步语音识别,请发出 POST 请求并提供相应的请求正文。以下示例展示了一个使用 curl 发出的 POST 请求。该示例使用通过 Google Cloud Cloud SDK 为项目设置的服务帐号的访问令牌。如需了解有关安装 Cloud SDK、建立项目和服务帐号以及获取访问令牌的说明,请参阅快速入门

以下示例展示了如何使用 curl 发送 POST 请求,其中请求正文启用了字词级置信度。

curl -s -H "Content-Type: application/json" \
    -H "Authorization: Bearer $(gcloud auth application-default print-access-token)" \
    https://speech.googleapis.com/v1p1beta1/speech:recognize \
    --data '{
    "config": {
        "encoding": "FLAC",
        "sampleRateHertz": 16000,
        "languageCode": "en-US",
        "enableWordTimeOffsets": true,
        "enableWordConfidence": true
    },
    "audio": {
        "uri": "gs://cloud-samples-tests/speech/brooklyn.flac"
    }
}' > word-level-confidence.txt

如果请求成功,服务器将返回一个 200 OK HTTP 状态代码以及 JSON 格式的响应(该响应会保存到名为 word-level-confidence.txt 的文件中)。

{
  "results": [
    {
      "alternatives": [
        {
          "transcript": "how old is the Brooklyn Bridge",
          "confidence": 0.98360395,
          "words": [
            {
              "startTime": "0s",
              "endTime": "0.300s",
              "word": "how",
              "confidence": 0.98762906
            },
            {
              "startTime": "0.300s",
              "endTime": "0.600s",
              "word": "old",
              "confidence": 0.96929157
            },
            {
              "startTime": "0.600s",
              "endTime": "0.800s",
              "word": "is",
              "confidence": 0.98271006
            },
            {
              "startTime": "0.800s",
              "endTime": "0.900s",
              "word": "the",
              "confidence": 0.98271006
            },
            {
              "startTime": "0.900s",
              "endTime": "1.100s",
              "word": "Brooklyn",
              "confidence": 0.98762906
            },
            {
              "startTime": "1.100s",
              "endTime": "1.500s",
              "word": "Bridge",
              "confidence": 0.98762906
            }
          ]
        }
      ],
      "languageCode": "en-us"
    }
  ]
}

Java

/**
 * Transcribe a local audio file with word level confidence
 *
 * @param fileName the path to the local audio file
 */
public static void transcribeWordLevelConfidence(String fileName) throws Exception {
  Path path = Paths.get(fileName);
  byte[] content = Files.readAllBytes(path);

  try (SpeechClient speechClient = SpeechClient.create()) {
    RecognitionAudio recognitionAudio =
        RecognitionAudio.newBuilder().setContent(ByteString.copyFrom(content)).build();
    // Configure request to enable word level confidence
    RecognitionConfig config =
        RecognitionConfig.newBuilder()
            .setEncoding(AudioEncoding.LINEAR16)
            .setSampleRateHertz(16000)
            .setLanguageCode("en-US")
            .setEnableWordConfidence(true)
            .build();
    // Perform the transcription request
    RecognizeResponse recognizeResponse = speechClient.recognize(config, recognitionAudio);

    // Print out the results
    for (SpeechRecognitionResult result : recognizeResponse.getResultsList()) {
      // There can be several alternative transcripts for a given chunk of speech. Just use the
      // first (most likely) one here.
      SpeechRecognitionAlternative alternative = result.getAlternatives(0);
      System.out.format("Transcript : %s\n", alternative.getTranscript());
      System.out.format(
          "First Word and Confidence : %s %s \n",
          alternative.getWords(0).getWord(), alternative.getWords(0).getConfidence());
    }
  }
}

Node.js

const fs = require('fs');

// Imports the Google Cloud client library
const speech = require('@google-cloud/speech').v1p1beta1;

// Creates a client
const client = new speech.SpeechClient();

/**
 * TODO(developer): Uncomment the following lines before running the sample.
 */
// const fileName = 'Local path to audio file, e.g. /path/to/audio.raw';

const config = {
  encoding: 'FLAC',
  sampleRateHertz: 16000,
  languageCode: 'en-US',
  enableWordConfidence: true,
};

const audio = {
  content: fs.readFileSync(fileName).toString('base64'),
};

const request = {
  config: config,
  audio: audio,
};

const [response] = await client.recognize(request);
const transcription = response.results
  .map(result => result.alternatives[0].transcript)
  .join('\n');
const confidence = response.results
  .map(result => result.alternatives[0].confidence)
  .join('\n');
console.log(`Transcription: ${transcription} \n Confidence: ${confidence}`);

console.log('Word-Level-Confidence:');
const words = response.results.map(result => result.alternatives[0]);
words[0].words.forEach(a => {
  console.log(` word: ${a.word}, confidence: ${a.confidence}`);
});

Python

from google.cloud import speech_v1p1beta1
import io

def sample_recognize(local_file_path):
    """
    Print confidence level for individual words in a transcription of a short audio
    file.

    Args:
      local_file_path Path to local audio file, e.g. /path/audio.wav
    """

    client = speech_v1p1beta1.SpeechClient()

    # local_file_path = 'resources/brooklyn_bridge.flac'

    # When enabled, the first result returned by the API will include a list
    # of words and the confidence level for each of those words.
    enable_word_confidence = True

    # The language of the supplied audio
    language_code = "en-US"
    config = {
        "enable_word_confidence": enable_word_confidence,
        "language_code": language_code,
    }
    with io.open(local_file_path, "rb") as f:
        content = f.read()
    audio = {"content": content}

    response = client.recognize(config, audio)
    # The first result includes confidence levels per word
    result = response.results[0]
    # First alternative is the most probable result
    alternative = result.alternatives[0]
    print(u"Transcript: {}".format(alternative.transcript))
    # Print the confidence level of each word
    for word in alternative.words:
        print(u"Word: {}".format(word.word))
        print(u"Confidence: {}".format(word.confidence))

使用远程文件

Java

/**
 * Transcribe a remote audio file with word level confidence
 *
 * @param gcsUri path to the remote audio file
 */
public static void transcribeWordLevelConfidenceGcs(String gcsUri) throws Exception {
  try (SpeechClient speechClient = SpeechClient.create()) {

    // Configure request to enable word level confidence
    RecognitionConfig config =
        RecognitionConfig.newBuilder()
            .setEncoding(AudioEncoding.FLAC)
            .setSampleRateHertz(44100)
            .setLanguageCode("en-US")
            .setEnableWordConfidence(true)
            .build();

    // Set the remote path for the audio file
    RecognitionAudio audio = RecognitionAudio.newBuilder().setUri(gcsUri).build();

    // Use non-blocking call for getting file transcription
    OperationFuture<LongRunningRecognizeResponse, LongRunningRecognizeMetadata> response =
        speechClient.longRunningRecognizeAsync(config, audio);

    while (!response.isDone()) {
      System.out.println("Waiting for response...");
      Thread.sleep(10000);
    }
    // Just print the first result here.
    SpeechRecognitionResult result = response.get().getResultsList().get(0);

    // There can be several alternative transcripts for a given chunk of speech. Just use the
    // first (most likely) one here.
    SpeechRecognitionAlternative alternative = result.getAlternativesList().get(0);
    // Print out the result
    System.out.printf("Transcript : %s\n", alternative.getTranscript());
    System.out.format(
        "First Word and Confidence : %s %s \n",
        alternative.getWords(0).getWord(), alternative.getWords(0).getConfidence());
  }
}

Node.js

// Imports the Google Cloud client library
const speech = require('@google-cloud/speech').v1p1beta1;

// Creates a client
const client = new speech.SpeechClient();

/**
 * TODO(developer): Uncomment the following line before running the sample.
 */
// const uri = path to GCS audio file e.g. `gs:/bucket/audio.wav`;

const config = {
  encoding: 'FLAC',
  sampleRateHertz: 16000,
  languageCode: 'en-US',
  enableWordConfidence: true,
};

const audio = {
  uri: gcsUri,
};

const request = {
  config: config,
  audio: audio,
};

const [response] = await client.recognize(request);
const transcription = response.results
  .map(result => result.alternatives[0].transcript)
  .join('\n');
const confidence = response.results
  .map(result => result.alternatives[0].confidence)
  .join('\n');
console.log(`Transcription: ${transcription} \n Confidence: ${confidence}`);

console.log('Word-Level-Confidence:');
const words = response.results.map(result => result.alternatives[0]);
words[0].words.forEach(a => {
  console.log(` word: ${a.word}, confidence: ${a.confidence}`);
});