Skip to main content
POST
Re-score Transcription Accuracy

Authorizations

Authorization
string
header
required

Use Tuner API key (tr_api_...) or user session token. Find your API key in Workspace Settings > API Keys.

Path Parameters

call_id
integer
required

Tuner's internal numeric ID for the call.

agent_id
integer
required

Tuner's internal numeric ID for the agent.

workspace_id
integer
required

Workspace ID. Find this in Workspace > General Settings.

Response

Successful Response

How accurately the agent's speech-to-text heard the replayed caller clips.

The score fields are stored when the call was scored. The verification fields, clips and is_outdated are read live from the clip set. Absent (null) on the parent when the call has no score.

score
number
required

Accuracy from 0 to 1 (1 minus the error rate, floored at 0)

error_rate
number
required

Raw word error rate. Can exceed 1 when much noise was transcribed

band
enum<string>
required

Quality band for the score

Available options:
good,
fair,
poor
reference_word_count
integer
required

Words the caller said (after cleaning)

hypothesis_word_count
integer
required

Words the agent heard (after cleaning)

wrong_count
integer
required

Caller words that were misheard

missed_count
integer
required

Caller words that were not transcribed

extra_count
integer
required

Transcribed words the caller did not say

error_entry_count
integer
required

Total number of mismatch entries, including those beyond the 50 stored

language
string
required

Language code the comparison used

partial_playback
boolean
required

True when the call ended before every clip was played

clips_total
integer
required

Clips in the set when the call was scored

clips_scored
integer
required

Clips that were played and compared

clips_verified
integer
required

How many of the scored clips are human-verified right now

is_outdated
boolean
required

True when clip transcripts were edited or verified after this call was scored, or the scoring rules changed. Re-score the call to use the current transcripts. False when the clip set no longer exists

calculated_at
string<date-time>
required

When the score was calculated

errors
TranscriptionErrorEntry · object[]

Up to 50 mismatches in reading order. See error_entry_count for the total

stt_provider
string | null

Raw speech-to-text provider name. Null when not reported

stt_model
string | null

Raw speech-to-text model name. Null when not reported

last_verified_at
string<date-time> | null

Latest verification time among the verified scored clips

clips
TranscriptionClipReference · object[]

The scored clips in playback order, with their current text