Transcripts and Audio
U-Analyze fetches the transcript payload from the Kafka-topic “contact-transcripts“ and the audio file from the configured shared path. It then uses this payload to process the audio and the evaluation of Business Rules, Scores, and Sentiments based on the category configuration. Reprocessing of this type of payload includes audio reprocessing along with re-evaluation of Business Rules and Scores other than transcription of audio calls.
Caveats
There is no differentiation in the UI to check whether the call is processed from U-Assist input
There is no UI restriction on the Reprocessing screen for U-Assist payloads
U-Assist redacts the PCI and PII fields at the transcript level and persists in Mongo DB. This is not done by U-Analyze.
Applicable for voice calls only and not for chat and email.
The voice call needs to be a single stereo file in .wav format and U-Analyze can process either the transcript alone or both the transcript and audio file.
Audio Redaction, and Signal Analytics are available only for transcripts with audio calls.
Both the U-Assist and U-Analyze audio file path must be located in the common network/server/mount path
Does not support processing of non-speech metadata
System property must be configured to identify Agent as first speaker or second based on the configuration given in U-Assis
If audio is not available, the transcript flow is taken for processing
Reprocessing of audio file from U-Assist in U-Analyze doesn't use the Transcriber module
Live and offline modes for voice calls are supported. (Chat and email are not supported).