単語レベルの信頼度の有効化

Speech-to-Text では、音声文字変換の単語ごとに正確さの値(信頼度レベル)を示すように指定できます。

単語レベルの信頼度

Speech-to-Text では、音声クリップの音声文字変換を行うとき、精度も測定してレスポンスします。Speech-to-Text から送信されるレスポンスでは、音声文字変換リクエスト全体の信頼度レベルが 0.0 から 1.0 の範囲の数値で示されます。 次のコードサンプルは、Speech-to-Text から返される信頼度レベルの値を示しています。

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

Speech-to-Text では、音声文字変換の全体の信頼度レベルに加えて、変換した個々の単語について信頼度レベルを示すこともできます。その場合、次の例に示すように、音声文字変換での WordInfo の詳細情報がレスポンスに含まれるようになり、単語ごとの信頼度レベルが示されます。

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

リクエストで単語レベルの信頼度を有効にする

次のコード スニペットは、ローカル ファイルとリモート ファイルを使用して Speech-to-Text への音声文字変換における単語レベルの信頼度を有効にする方法を示しています。

ローカル ファイルの使用

プロトコル

詳細については、speech:recognize API エンドポイントをご覧ください。

同期音声認識を実行するには、POST リクエストを作成し、適切なリクエスト本文を指定します。次は、curl を使用した POST リクエストの例です。この例では、Google Cloud Cloud SDK を使用して、プロジェクト用に設定されたサービス アカウントのアクセス トークンを扱います。Cloud SDK のインストール、サービス アカウントがあるプロジェクトの設定、アクセス トークンの取得などの手順については、クイックスタートをご覧ください。

次の例は、curl を使用して POST リクエストを送信する方法を示しています。ここでは、リクエストの本文で単語レベルの信頼度を有効にしています。

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

リクエストが成功すると、サーバーは 200 OK HTTP ステータス コードと JSON 形式のレスポンス(word-level-confidence.txt ファイルに保存される)を返します。

{
  "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))

リモート ファイルの使用

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}`);
});