TheRealKevinGlinski | 2020-11-11 14:27:35 UTC | #1
When downloading call transcripts with /api/v2/speechandtextanalytics/conversations/, each conversation has a list of phrases. How does the transcription service define a phrase? ie where is the boundary between one phrase and the next? If I join all the phrases together, will that give me the entire conversation, or just the phrases that met a certain confidence level?
shahar.mosek | 2020-11-11 14:43:18 UTC | #2
The basic definition is by pauses. A long enough pause will generally end the phrase. There is no filtering of phrases by confidence, but only words that pass the defined confidence threshold are shown. So joining all phrases together will give you the entire conversation that was identified with a high enough confidence.
TheRealKevinGlinski | 2020-11-11 15:00:11 UTC | #3
Awesome, thanks for the confirmation
system | 2020-12-12 15:00:12 UTC | #4
This topic was automatically closed 31 days after the last reply. New replies are no longer allowed.
This post was migrated from the old Developer Forum.
ref: 9287