Sentiment Domain

Overview#

A 911 supervisor reviewing a call receives an automatic annotation: high panic indicator, elevated speech rate, keyword match on "weapon." The call is already being handled, but the annotation helps the supervisor understand what the dispatcher was responding to. Later in the shift, a wellness dashboard shows that one call-taker has received an unusually high proportion of high-stress calls in the last two hours. The Sentiment domain is what produces both of those signals: AI-powered analysis of communications that detects emotional markers, speech pattern shifts, and critical keywords, then surfaces them as risk scores, priority recommendations, and trend data.

Key Features#

• Emotional Analysis: Detect emotional indicators including panic, fear, anger, and deception in communications with confidence scoring to help supervisors understand the emotional context of interactions.

• Keyword Detection: Identify critical keywords and phrases in communications that may indicate the severity or nature of an emergency, supporting faster and more accurate categorisation.

• Risk Scoring: Aggregate emotional indicators, keyword matches, and communication patterns into an overall risk score that helps prioritise response decisions.

• Priority Recommendations: Generate AI-powered priority recommendations based on sentiment analysis results to assist dispatchers in routing calls during high-volume periods.

• Speech Pattern Analysis: Evaluate communication patterns including speech rate and volume changes to provide additional context about caller or operator state.

• Batch Processing: Analyse multiple communications in a single operation for efficient processing of historical data and shift-level sentiment reviews.

Use Cases#

Sentiment analysis of communications has direct value wherever call volume is high and human reviewers cannot monitor every interaction in real time. Relevant industries include public safety and emergency services, healthcare call centres, and financial services customer operations.

• Emergency Call Triage: Automatically assess the urgency of incoming calls based on caller sentiment, keyword detection, and speech patterns to support dispatcher prioritisation decisions.

• Operator Wellness Monitoring: Track sentiment patterns across dispatchers and call-takers over time to identify signs of burnout or elevated stress before they affect performance.

• Quality Assurance: Review sentiment trends across shifts and teams to identify areas where additional training or procedural changes could improve service quality.

• Post-Incident Review: Analyse communications from completed incidents to evaluate emotional dynamics and identify opportunities for improved response protocols.

Integration#

The Sentiment domain enhances communication management across the platform:

• Communication Management: Sentiment analysis enriches call and message records
• Workforce Management: Stress indicators feed into operator wellness monitoring
• Analytics: Sentiment trends contribute to operational dashboards
• Alert System: Elevated urgency or distress levels can trigger supervisor notifications

Open Standards#

• GraphQL (June 2018 specification): the entire Sentiment domain API surface is exposed as a typed GraphQL schema, with named queries (sentimentAnalysis, sentimentAnalysesByCall, sentimentStatistics) and a mutation (analyzeSentiment) conforming to the GraphQL specification.
• RFC 7519 (JSON Web Token): all GraphQL operations in the domain are gated by JWT-based authentication; the IsAuthenticated permission class validates bearer tokens issued by the platform identity service before any analysis data is read or written.
• NENA i3 / NG9-1-1 (NENA-STA-010): the priority bands P1, P5 produced by the risk-scoring algorithm, and the PSAP-specific emotional categories (calm, anxious, distressed, panicked, aggressive), align with the NENA i3 emergency call priority and triage vocabulary used across the platform's NG9-1-1 workstreams.
• APCO ANS 1.102.1 (Public Safety Communications): the call-taker wellness monitoring and call quality evaluation dimensions (speech rate, pause count, operator stress indicators) follow APCO/NENA joint standards for dispatcher performance and wellbeing assessment in public-safety communications centres.
• ISO 8601: all analysis timestamps (analyzed_at, created_at) are stored and returned as UTC-offset datetime values, conforming to ISO 8601 date-time interchange format.
• JSON: sentiment results are serialised as JSON via Pydantic models and transmitted over the GraphQL transport layer, using JSON as the canonical data interchange format for all analysis payloads.

Last Reviewed: 2026-02-05 Last Updated: 2026-04-14