完整的听录显示在 text 属性中。 你可以查看整个单词、音节和特定音素的准确度分数。 可使用语音 SDK 获得相同的结果。 有关信息,请参阅如何使用发音评估。
{
"text": "Today was a beautiful day. We had a great time taking a long long walk in the morning. The countryside was in full bloom, yet the air was crisp and cold towards end of the day clouds came in forecasting much needed rain.",
"duration": 156100000,
"offset": 800000,
"json": {
"Id": "f583d7588c89425d8fce76686c11ed12",
"RecognitionStatus": 0,
"Offset": 800000,
"Duration": 156100000,
"DisplayText": "Today was a beautiful day. We had a great time taking a long long walk in the morning. The countryside was in full bloom, yet the air was crisp and cold towards end of the day clouds came in forecasting much needed rain.",
"SNR": 40.47014,
"NBest": [
{
"Confidence": 0.97532314,
"Lexical": "today was a beautiful day we had a great time taking a long long walk in the morning the countryside was in full bloom yet the air was crisp and cold towards end of the day clouds came in forecasting much needed rain",
"ITN": "today was a beautiful day we had a great time taking a long long walk in the morning the countryside was in full bloom yet the air was crisp and cold towards end of the day clouds came in forecasting much needed rain",
"MaskedITN": "today was a beautiful day we had a great time taking a long long walk in the morning the countryside was in full bloom yet the air was crisp and cold towards end of the day clouds came in forecasting much needed rain",
"Display": "Today was a beautiful day. We had a great time taking a long long walk in the morning. The countryside was in full bloom, yet the air was crisp and cold towards end of the day clouds came in forecasting much needed rain.",
"PronunciationAssessment": {
"AccuracyScore": 92,
"FluencyScore": 81,
"CompletenessScore": 93,
"PronScore": 85.6
},
"Words": [
// Words preceding "countryside" are omitted for brevity...
{
"Word": "countryside",
"Offset": 66200000,
"Duration": 7900000,
"PronunciationAssessment": {
"AccuracyScore": 30,
"ErrorType": "Mispronunciation"
},
"Syllables": [
{
"Syllable": "kahn",
"PronunciationAssessment": {
"AccuracyScore": 3
},
"Offset": 66200000,
"Duration": 2700000
},
{
"Syllable": "triy",
"PronunciationAssessment": {
"AccuracyScore": 19
},
"Offset": 69000000,
"Duration": 1100000
},
{
"Syllable": "sayd",
"PronunciationAssessment": {
"AccuracyScore": 51
},
"Offset": 70200000,
"Duration": 3900000
}
],
"Phonemes": [
{
"Phoneme": "k",
"PronunciationAssessment": {
"AccuracyScore": 0
},
"Offset": 66200000,
"Duration": 900000
},
{
"Phoneme": "ah",
"PronunciationAssessment": {
"AccuracyScore": 0
},
"Offset": 67200000,
"Duration": 1000000
},
{
"Phoneme": "n",
"PronunciationAssessment": {
"AccuracyScore": 11
},
"Offset": 68300000,
"Duration": 600000
},
{
"Phoneme": "t",
"PronunciationAssessment": {
"AccuracyScore": 16
},
"Offset": 69000000,
"Duration": 300000
},
{
"Phoneme": "r",
"PronunciationAssessment": {
"AccuracyScore": 27
},
"Offset": 69400000,
"Duration": 300000
},
{
"Phoneme": "iy",
"PronunciationAssessment": {
"AccuracyScore": 15
},
"Offset": 69800000,
"Duration": 300000
},
{
"Phoneme": "s",
"PronunciationAssessment": {
"AccuracyScore": 26
},
"Offset": 70200000,
"Duration": 1700000
},
{
"Phoneme": "ay",
"PronunciationAssessment": {
"AccuracyScore": 56
},
"Offset": 72000000,
"Duration": 1300000
},
{
"Phoneme": "d",
"PronunciationAssessment": {
"AccuracyScore": 100
},
"Offset": 73400000,
"Duration": 700000
}
]
},
// Words following "countryside" are omitted for brevity...
]
}
]
}
}
发音评估支持不间断的流式处理模式。 AI Studio 演示允许在流式处理模式下进行长达 60 分钟的录制,以便进行评估。 只要不按下停止录制按钮,评估过程就不会结束,你可以方便地暂停和恢复评估。
发音评估对发音的几个方面进行了评估。 在“评估结果”的底部,可以看到“发音分数”显示为综合整体分数,它包含 4 个子方面:“准确度分数”、“流畅度分数”、“完整性分数”和“韵律分数”。 在流式处理模式下,由于准确度分数、流畅性分数和韵律分数在整个录制过程中会随时间而变化,因此我们在 AI Studio 上演示了这样一种方法:在评估结束前逐步显示大致整体分数,该分数只使用准确度分数、流畅性分数和韵律分数进行加权。 完整性分数只在评估结束时你按下停止按钮后进行计算,因此最终的发音整体分数根据准确度分数、流畅性分数、完整性分数和韵律分数加权汇总得出。