Skip to main content

Speech Insights

This section covers the WinnerWare modules used to collect call recordings, turn them into redacted and searchable transcripts, apply AI tags and indicators, and support review, export, and queue-based operations.

Overview​

Speech Insights is a distinct WinnerWare capability area focused on call transcript operations and analytics. It is separate from the base Speech modules, which focus on voice playback and speech-service providers.

A call moves through Speech Insights in this order:

  1. Ingestion - WinnerWare collects the recording and its metadata from an ingestion source.
  2. Transcription - Azure AI Speech turns the audio into text with speakers and timing.
  3. Redaction - AI masks sensitive values in the text, the search index, and the audio.
  4. Tags and indicators - WinnerWare classifies the call and extracts values from it.
  5. Search and review - Reviewers search, open, listen to, and export calls.
  6. Retro processing - New or changed tags and indicators are applied to calls that were already processed.

These modules are typically used together:

  • Speech Insights provides the core admin, search, tagging, redaction, retro processing, export, and queue experience
  • Speech Insights FTP adds FTP/FTPS-based ingestion sources
  • Speech Insights SFTP adds SFTP-based ingestion sources

Modules in this section​

  • Speech Insights - Describes the full flow from ingestion to search, including ingestion sources, redaction, tags, indicators, the call page, exports, and the processing queue.
  • Speech Insights Retro Processing - Describes how to apply tags and indicators to older calls, run statuses, and what to do when a run fails.
  • Speech Insights Expression Search - Describes the Lucene-lite speech analytics search syntax used in Expression mode and in Expression tags.
  • Speech Insights FTP - Adds FTP and FTPS ingestion for recording source locations.
  • Speech Insights SFTP - Adds SFTP ingestion for recording source locations.