📓 SpeechEvalPro API

Public Parameters

1. Evaluation parameters #

StartMessage #

common field content is fixed

payload field have required parameters and optional parameters

payload.params Refer to the introduction of each question type

field type remark defaults
langType enum english：en-US
chinese：zh-cmn-Hans-CN		 required
connectTimeout int Connection 、Heartbeat timeout (seconds), range: 5-60. 15
responseTimeout int Response timeout (seconds), range: 5-60. 15
rank int Scoring scale, range: 1-100 100
precision float Score ad:justment coefficient, range: 0.8-1.5 (up to 3 decimal places)
The function of this coefficient is to adjust up or down the score between 30 and 90 (percentage system), and the scores in other ranges remain unchanged
When precision >1.0, the score is raised, the closer to 75 points, the higher the adjustment degree
When precision <1.0, the score is lowered, the closer to 75 points, the higher the adjustment degree
When precision =1.0, no score adjustment		 1.0
attachAudioUrl boolean Deprecated Whether to return the audio download URL
The audio is kept for about 10 days by default. If you need to save it permanently, it is recommended that the access party download it to its own server		 false
realtime boolean Whether to return the result in real time
Only Chinese/English sentence, chapter question types are supported.		 false

Examples

{
    "common": {
        "api":"speecheval",
        "cmd": "start"
    },
    "payload": {
        "langType": "en-US",
        "format": "pcm",
        "params": {
            "refText": "Yes(t:1)?",
            "mode": "sentence"
        }
    }
}

StopMessage # 

{
    "common": {
        "cmd": "stop",
        "api": "speecheval"
    }
}

2. Public parameters in the response #

audioInfo（Audio Quality Information）field #

params type remark
snr double Signal-to-noise precision, the higher the value, the clearer it is (-20～+60dB)
When this parameter affects the score, the corresponding tipId will be set
clip double The volume is too loud, truncation occurs (0~1)
When this parameter affects the score, the corresponding tipId will be set
volume double Recording volume (0~90dB)
tipId string Audio Quality Type ID

tipId introduction

tipId description remark
00000 normal audio None
90001 The user’s pronunciation is incomplete, such as “good morning”, may only say “good” Can prompt incomplete pronunciation
90002 The recognition is incomplete. In this case, most of them are silent and the audio is short. Can prompt incomplete pronunciation
90003 Low volume, probably located too far away Can advise users to adjust microphone distance or volume
90004 Audio clipped, maybe too close Can advise users to adjust microphone distance or volume
90005 Poor audio quality (low signal-to-noise precision caused by noisy recording environment or indistinct speech) It can prompt the user that the voice is not obvious