The client_data object

The top-level client_data object contains all custom metadata fields and values submitted with the initial ASR transcription request. The client_data object also includes the following fields created and populated by the system:

Important: The client_data object is generated by V‑Spark, and does not appear in V‑Blaze or V‑Cloud JSON transcripts.
Table 1. Elements in the client_data object created by V‑Spark

Element

Type

Description

audio_duration

number

The initial audio file's duration in seconds.

audio_properties

string

The initial audio's file format, encoding type, bitrate, frequency, and number of channels. These properties may not be available for all audio. In that case, the comma-separated list of values will include blanks, as in the following example:

{...
"audio_properties": "Opus, , , 48000 Hz, , 2 ch",
...}

jm_version

string

The version of Job Manager that processed the transcript. As of V‑Spark 4.0, this field always has the value "None" because Job Manager version now matches the V‑Spark version.

performance_stats

object

Stores the following metrics from ASR processing and transcript analysis:

job_latency

number

Represents the efficiency of the analysis job as a function of end time, start time, and audio duration as in the formula (EndTime - StartTime) / (AudioDuration) . In other words, a measure of how long it took to process the transcript in comparison to the length of the audio file.

job_end

string

The time at which the job ended.

asr_latency

number

Represents the efficiency of the ASR engine's transcription job as a function of end time, start time, and audio duration as in the formula (EndTime - StartTime) / (AudioDuration) . In other words, a measure of how long it took to transcribe the audio in comparison to the length of the audio file.

asr_end

string

The time at which ASR transcription ended.

job_start

string

The time at which the job began.

asr_start

string

The time at which ASR transcription began.

Custom metadata submitted with the initial audio file is always provided in JSON output exactly as submitted, but custom metadata is indexed and searchable from the UI only if the metadata field is configured for the audio's V‑Spark folder when that audio is processed. Custom metadata that is submitted with the audio but not configured for the processing folder is passed through to the client_data object in JSON output, but that metadata is neither indexed nor searchable.

The following is an example of the client_data object in a JSON transcript:

"client_data": {
    "agentname": "Reid Smith",
    "client-phone": "123-456-7890",
    "branch-loc": "Raleigh",
    "account_priority": "Medium",
    "jm_version": "None",
    "audio_properties": "AAC, CBR, 175 kb/s, 48000 Hz, , 2 ch",
    "performance_stats": {
      "job_latency": 0.4851,
      "job_end": "2021-10-06_19:22:41 +0000",
      "asr_latency": 0.1541,
      "asr_end": "2021-10-06_19:22:40 +0000",
      "job_start": "2021-10-06_19:22:39 +0000",
      "asr_start": "2021-10-06_19:22:39 +0000"
    },
    "audio_duration": 4.1387
  }

Custom metadata can be provided in either JSON or XML format. The preceding example client_data section was produced in response to uploading the following custom metadata JSON in a file zipped with a corresponding audio file:

{
	"metadata": {
	  "agentname": "Reid Smith",
	  "client-phone": "123-456-7890",
	  "branch-loc": "Raleigh",
	  "account_priority": "Medium"
	}
}

The following example shows the XML equivalent of the preceding JSON metadata file:

<metadata>
	<agentname>Reid Smith</agentname>
	<client-phone>123-456-7890</client-phone>
	<branch-loc>Raleigh</branch-loc>
	<account_priority>Medium</account_priority>
</metadata>