Get corrected transcripts

Your API key for authentication.

The unique identifier for the call to be corrected.

Will be success if the request was successful.

Confirms the request was successful, or provides an error message if the request failed.

This will contain an array of objects. Each object will be constructed as the following.

{
"start": 0.069, // start time of the transcript
"end": 2.551, // end time of the transcript
"text": " Hi, I'm calling about a pizza order.", // the corrected text
"speaker": 0 // the identified speaker diarization. Can be 0,1,2,3 etc
}

Corrected transcripts provides us with a raw output that is generally unusable because we can’t eveen neccessarily align the ‘assistant’ and ‘user’ roles. To fix this, we provide our version of an ‘aligned’ transcript. This means essentailly a transcript where the roles are matched to the pieces of text.

We do this by vectorizing the text, taking the cosine similarity, and adding a predictive layer based off of the wait_for_greeting param (essentially how sure are we that assistant or user spoke first).

This will contain an array of objects. Each object will be constructed as the following.

{
"id": 3056004,
"created_at": "2024-02-29T18:40:41.26799+00:00",
"text": "Great, Thanks John. Could you tell me about the pizza order you placed?", // the corrected text
"user": "assistant", // the presumed role
"c_id": "bfaf99a1-b7c0-4f96-9630-90bc41cea488"
},