Skip to main content

Speech Insights

Speech Insights is the WinnerWare module for turning call recordings and completed transcripts into searchable, governable, and reportable insight.

Overview

This module gives teams one place to manage the full post-call workflow. It helps WinnerWare move from raw recordings to structured transcript insight by combining ingestion, processing, search, tagging, redaction, saved lookups, export, and queue management.

Key Features

  • Provides the main Speech Insights admin area in WinnerWare.
  • Supports multiple ingestion sources for bringing recordings into the platform.
  • Supports transcript tags, indicators, and saved lookups for structured review.
  • Supports sensitive-data redaction rules for safer transcript handling.
  • Supports a processing queue so teams can monitor and retry transcript work.
  • Supports export settings and downloadable transcript exports.
  • Supports API-style access to indexed transcript data when that is needed for connected tools.

How It Works

Bring recordings in through managed ingestion sources

Speech Insights starts with an ingestion source. Teams create source definitions for the locations where call recordings and their metadata are delivered, then activate the sources that should participate in processing.

Each source includes:

  • a business-facing name
  • the related client
  • the locale for the incoming records
  • an active or inactive state
  • a field map that explains how source metadata should be interpreted

The field map is important because it tells WinnerWare which incoming fields represent the recording file name, call date, agent information, queue information, and other call metadata. At minimum, the source must map the audio file name and call date.

Process completed transcripts into usable insight

After a recording is ingested, WinnerWare moves it through transcript processing and enriches the result with insight-oriented structures such as tags, indicators, and redaction logic. This is what turns a raw transcript into something teams can search, review, govern, and reuse.

Search and export by transcript index

Speech Insights creates search experiences around transcript indexes. Once an index is available, teams can open the related Speech Insights search page, choose between Keyword, Semantic, and Expression search modes, review completed transcript matches, and export results when the export feature is enabled.

Monitor operational health with the processing queue

The Processing Queue gives administrators a way to see pending, processing, completed, and abandoned work. Abandoned items can be re-queued, which helps teams keep transcript processing moving without manual data repair.

Configuration

Enable the features your rollout needs

Admin display nameFeature IDUse when
Speech InsightsCloudSolutions.Speech.InsightsYou need the main Speech Insights admin, search, tagging, redaction, saved lookup, and queue features.
Filesystem Recording Ingestion SourceCloudSolutions.Speech.Insights.FilesystemYou want recordings to be ingested from a managed tenant file location.
Speech Insights ExportCloudSolutions.Speech.Insights.ExportYou want downloadable transcript exports and export retention settings.
Speech Insights Export - AzureCloudSolutions.Speech.Insights.Export.AzureYou want exported files stored in Azure instead of the local tenant file system.

Main admin areas

The module adds these main admin surfaces:

Admin areaPurpose
Speech Insights -> Search: Index NameSearch completed transcripts for a speech index profile.
Speech Insights -> Ingestion SourcesCreate and manage recording source connections.
Speech Insights -> Transcript TagsMaintain tag definitions used in transcript review.
Speech Insights -> IndicatorsMaintain indicator definitions used for call insight extraction.
Speech Insights -> Redaction RulesControl how sensitive content should be masked.
Speech Insights -> Saved LookupsSave reusable query patterns for repeated review work.
Speech Insights -> Processing QueueMonitor queue health and retry abandoned items.

Permissions

PermissionPurpose
Manage Ingestion SourcesCreate, edit, run, and delete ingestion sources.
Manage Speech Transcript TagsMaintain transcript-tag definitions.
Manage Speech IndicatorsMaintain indicator definitions.
Manage Speech LookupsMaintain saved lookup definitions.
Manage Sensitive Data Redaction RulesMaintain masking and redaction rules.
Manage speech transcript processing queueMonitor queue health and retry abandoned work.
Manage Speech Insights SettingsEdit the Speech Insights export settings page.
Export speech transcriptGenerate and download transcript exports.
Access indexed data using APIAccess indexed transcript data through the API-style OData layer.

Export settings

When the export feature is enabled, WinnerWare adds one export setting:

  • Report retention (days) - controls how long exported reports remain available for download

Usage

Build a practical call-review workflow

A common rollout starts by defining one or more ingestion sources, then adding the tags, indicators, and redaction rules that fit the review process. Once transcripts are flowing, teams can work from the search screens, saved lookups, and exports to support review and reporting.

Pick the right search mode

Speech Insights offers three complementary search styles:

  • Keyword for direct term and filter-based lookups
  • Semantic for concept-driven retrieval using embeddings
  • Expression for structured transcript logic such as boolean groups, proximity, speaker targeting, early-call windows, pause-aware phrases, and tag requirements

Example setup

One clear setup pattern is:

  1. enable the main Speech Insights feature and the source or export features the tenant needs
  2. create an ingestion source and map the audio file name and call date fields first
  3. add transcript tags, indicators, and redaction rules that match the review process
  4. let recordings process, then use the search area and processing queue to confirm data is flowing correctly
  5. add saved lookups or exports once the business review pattern is stable

Use saved structures instead of ad hoc review

Speech Insights works best when teams standardize how they classify and review calls. Transcript tags, indicators, redaction rules, and saved lookups all help move the experience away from one-off searches and toward repeatable operating patterns.

Developer Notes

  • The module seeds default redaction rules during setup.
  • The export endpoint supports direct CSV output for smaller exports and background generation for larger result sets.
  • An OData-style endpoint is available under speech/insights/odata/{indexName} for connected tools that need indexed transcript data.
  • OData document access uses basic authentication and requires the Access indexed data using API permission.
  • Transcript tags, indicators, and redaction rules include structured import and deployment support, which helps teams move approved configurations between environments.