Location Intelligence: Multi-Source Address and Coordinate Resolution

Overview#

The Location Intelligence module resolves caller-stated location information, in any format, into precise geographic coordinates before emergency units are dispatched. Callers describe their location in dozens of ways: a street address, a what3words code shown on a sign, a Google Plus Code from a hiking app, a landmark or business name, or simply the name of a place they recognise. The module tries each resolution method in a prioritised waterfall, falling back to the caller's registered address if all other methods fail, and returns a confidence score with every result so dispatchers can immediately assess how precisely the location has been pinned.

A GPS mismatch detection layer cross-checks stated locations against any device-reported coordinates, flagging discrepancies that may indicate a false report or a caller who is in a different location than described.

Key Features#

what3words Resolution: Recognises the ///word.word.word format and resolves it to coordinates via the what3words API, delivering 3-metre square precision with a 95% confidence score, particularly valuable for rural incidents, events in open spaces, and locations without conventional postal addresses
Google Plus Code Resolution: Detects Plus Code patterns (such as 9C3W9QCJ+2V) and geocodes them via the Google Maps API with regional context for improved accuracy, returning a 90% confidence score
Street Address Geocoding: Full postal address text is geocoded against the Google Maps API, with confidence scoring tiered by result type: rooftop matches (95%), interpolated range (85%), geometric centre (75%), and approximate results (60%)
POI and Landmark Lookup: Location text that does not match an address or code format is treated as a point-of-interest or business name query, using fuzzy matching against the Google Places API with stopword stripping for improved hit rates; a two-pass approach tries strict matching before falling back to fuzzy
Caller-Registered Address Fallback: When all other resolution methods fail, the caller's pre-registered home or work address from their profile is used as a fallback, ensuring dispatch is never completely uninformed about location
Confidence-Scored Results: Every resolved location includes a numeric confidence value (0.0 to 1.0) and a source label (what3words, plus_code, google, poi, registered, unresolved), giving dispatchers transparent signal quality alongside the coordinate
GPS Mismatch Detection: When device GPS coordinates are available alongside a stated address, the module calculates the Haversine distance between the two. Discrepancies exceeding 5 kilometres are flagged as a mismatch and surfaced to the dispatcher
Graceful Degradation: If third-party location APIs are unavailable, the module returns a partial result at the deepest level reached rather than blocking dispatch; all external API keys are optional and the module operates in reduced-accuracy mode without them

Use Cases#

Rural Emergency with what3words#

A hiker injured on an unmarked trail shares the what3words code visible on a nearby trail marker sign. The system resolves it to a 3-metre square in under a second, giving ambulance crews an exact location they could not have obtained from a verbal description of the surrounding landscape.

Event Location via Plus Code#

A caller at an outdoor festival uses the Plus Code shown on the event programme. The system decodes and geocodes it, pinning the location to within a building footprint even though the caller does not know the full postal address of the venue.

Business Name Incident#

A caller says "I'm outside the Tesco on the Main Road in Kilkenny." POI lookup identifies the specific branch, resolves it to precise coordinates, and returns the address with an 80% confidence score for dispatcher review.

GPS Cross-Verification#

A caller claims to be calling from an address in one part of the city, but their device GPS places them 12 kilometres away in a different neighbourhood. The GPS mismatch flag is raised in the dispatcher interface, prompting a verification question before units are committed.

No Address Available#

An anonymous caller cannot or will not provide a location. All resolution attempts fail. The system falls back to the caller's registered home address if a profile exists, providing the dispatcher with a probable location rather than nothing.

How It Works#

Diagram

flowchart TD
    A[Caller States Location] --> B{what3words format?}
    B -->|Yes| C[what3words API → 95% confidence]
    B -->|No| D{Plus Code format?}
    D -->|Yes| E[Google Geocode API → 90% confidence]
    D -->|No| F{Address text?}
    F -->|Yes| G[Google Geocode\nROOFTOP 95% / RANGE 85% / CENTRE 75%]
    F -->|No| H{POI / landmark text?}
    H -->|Yes| I[Google Places\nStrict → Fuzzy → 70–90% confidence]
    H -->|No| J[Registered Caller Address Fallback]
    J --> K{Profile exists?}
    K -->|Yes| L[Return registered address]
    K -->|No| M[Unresolved, dispatcher manual entry]

    C --> N[LocationResult]
    E --> N
    G --> N
    I --> N
    L --> N

    N --> O{GPS coords available?}
    O -->|Yes| P{Distance > 5km?}
    P -->|Yes| Q[GPS Mismatch Flag → Dispatcher Alert]
    P -->|No| R[Location Confirmed]
    O -->|No| R

Integration#

Location Intelligence is invoked during call setup in the Voice AI session layer, and its resolved coordinates are passed directly to the Incident Management module for unit dispatch. The confidence score and source type are displayed in the dispatcher's call panel alongside the resolved address, enabling immediate human review of lower-confidence results. what3words and Google Maps API keys are configured at the organisation level through the administration console. The GPS mismatch flag is surfaced as an advisory event in the incident record and dispatcher interface, not as an automatic block on dispatch.

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

Module metadata