What processing does U-Discover perform on conversations?
When a conversation is processed by U-Discover, it means that it goes through a series of audio and text processing services to deliver key metrics and insights. You can process a batch of calls using Process Batch Conversation.
During processing, every conversation goes through the following audio and text processing services:
Automatic Speech Recognition (ASR): Transcribes the call and identifies speakers (Agent and Customer). U-Dis supports US English.
Data Redaction: Redacts PII data of the customer
Voice Activity Detection: Analyzes audio related activity of the call. All these can be viewed in the Conversation Detail section of each conversation in U-Discover. The following audio regions are identified:
Customer talk
Agent talk
Talkover
Silence overlap – only overlapping silence
Music - removed
Other processing:
Rate of speech – transcript no. of words of agent by agent talk duration
Call hold count – if both are silent at the same time it is identified as hold, and such overlapping silent regions are counted as one hold
Tonal sentiment detection: Analyzes both text and tone to identify customer sentiment throughout the call. Tonal sentiment is detected using Uniphore’s Emotion AI service.
Facts: Resolves Facts
In addition, Scorecards are evaluated against each call in the respective Organization(s) where the scorecards are created.