[Développeurs]

Multilingual Call-Taking with Live Translation and ISL Relay

A call-taker can answer a 999 call from a Polish-speaking caller and see both the Polish transcript and live English translation streaming simultaneously, with the translated text routed straight into the protocol-triage

Catégorie: GeospatialDernière Mise à Jour: 5 mai 2026
geospatialaireal-time

Overview#

A call-taker can answer a 999 call from a Polish-speaking caller and see both the Polish transcript and live English translation streaming simultaneously, with the translated text routed straight into the protocol-triage engine without losing a beat.

The Multilingual Call-Taking module brings real-time speech-to-text and translation into the PSAP active-call panel for any of the major EU languages plus Polish, Portuguese, Romanian, Lithuanian, Mandarin, Arabic, and Tigrinya. A parallel video relay path supports Irish Sign Language and International Sign callers via deaf-led interpreter pool, so accessibility is treated as a first-class call type rather than an afterthought.

Last Reviewed: 2026-05-05 Last Updated: 2026-05-05

Key Features#

  • Live Bilingual Transcript Stream: Source-language and English transcripts render side-by-side in the call-taker's active-call panel as the caller speaks.

  • Automatic Language Detection and Provider Routing: A multilingual router selects the best ASR provider per language and confidence profile rather than locking the PSAP to a single vendor.

  • Protocol Engine Continuity: The English translation is fed into the AMPDS protocol-triage engine so card-based questioning still functions for non-English callers.

  • ISL and International Sign Video Relay: A WebRTC route connects the caller to a deaf-led interpreter, so a sign-language 999 contact reaches the call-taker as live, signed-then-translated speech.

  • Dual-Language Persistence for Governance: Both the original-language and English transcripts are stored against the canonical incident with ISO 639-3 language tags so reviewers can audit either version.

  • Confidence and Provider Tagging: Each transcript segment carries the ASR provider id and a confidence score so reviewers can flag uncertain wording.

  • Reuse of the ePCR Translation Service: The same clinical translation service already trusted in ePCR is reused for PSAP, keeping translation behaviour consistent across the prehospital pathway.

  • AIHE Handoff in Working Language: The English transcript travels with the AIHE hospital handoff so the receiving ED reads context in their working clinical language.

Use Cases#

  • Polish-Speaking 999 Caller: A call-taker answers, sees Polish and English transcripts streaming together, and proceeds through AMPDS card questioning on the translated text.

  • Tigrinya-Speaking Caller in Distress: The router selects a multilingual ASR profile capable of Tigrinya and surfaces both transcripts so the call-taker can keep the caller talking while triage runs.

  • ISL Caller via Video Relay: A Deaf caller initiates a video relay session; a deaf-led interpreter joins, signs are voiced in English, and the call-taker sees a live transcript in the same panel.

  • International Sign for Visiting Caller: A caller using International Sign reaches the same interpreter pool, with the relayed speech transcribed and queued into the protocol engine.

  • Clinical Governance Review: A reviewer opens the canonical incident later and reads both the source-language transcript and the English translation, with confidence and provider id visible per segment.

  • Receiving ED Handoff: The ED clinician taking the AIHE handoff reads the English summary, knowing the source-language audio and transcript are preserved against the incident.

Integration#

  • PSAP Active-Call Panel: The Translator panel renders the dual-language transcript inside the existing call-handling UI rather than a separate window.

  • AMPDS Protocol Engine: The English translation stream is the input the triage engine consumes, so multilingual calls behave the same as English-language calls from the engine's perspective.

  • Canonical Incident Store: Source and translated transcripts are written as separate separate transcript records linked to the same incident, each tagged with an ISO 639-3 language code.

  • ePCR Translation Service: PSAP reuses the same translation service already in production for ePCR, keeping clinical translation behaviour aligned across call-taking and patient care.

  • AIHE Hospital Handoff: The English transcript flows into the AIHE hospital pre-alert payload so the receiving ED gets context in their working language.

  • Argus Responder Mobile: The TranslatorScreen pattern already used by field crews sets the UX baseline so call-takers and responders see translation in a familiar shape.

  • Multi-Provider ASR: The transcription provider layer routes to Deepgram or OpenAI Whisper based on language coverage, latency, and confidence profile.

  • Video Relay Infrastructure: A WebRTC binding connects callers to an interpreter pool so ISL and International Sign calls reach the same call-taker workflow.

Open Standards#

  • ISO 639-3: every transcript segment is tagged with a standard language code so downstream systems can filter and display by language.
  • W3C WebRTC: the video relay channel between caller, interpreter, and call-taker is built on the open browser-native real-time communications stack.
  • ETSI ES 202 975: multilingual emergency call handling expectations for PSAPs are reflected in how the module routes and presents non-native-language calls.
  • ETSI EN 301 549: accessibility requirements for ICT including emergency relay services shape the ISL and International Sign video relay path.
  • IETF RFC 3261 (SIP) and RFC 4103 (Real-Time Text): relay sessions use the SIP signalling and real-time text transport standards already adopted by modern relay services.
  • W3C WebVTT: live transcript captions are rendered using the standard web caption format expected by modern browsers and accessibility tooling.
  • CloudEvents 1.0: translation completion notifications are published using a common structured event envelope so other platform services can react to translated transcripts without tight coupling.

Prêt à Développer ?

Commencez avec nos APIs ou contactez notre équipe d'intégration pour obtenir de l'aide.