1 of 12

911 EMERGENCY AI ASSISTANT (EmAA)

  • Aishwarya Kumaran
  • Hiten Kothari
  • Sudarsana Jayandan Janakaraj

2 of 12

IMPACT OF 911 EMERGENCY CALLS

  • Estimated 240 million 911 call each year - 650,000 each day - 7.6 calls per second
  • Answered within 10 sec for 90% of calls
  • PSAPs were only able to achieve the minimum time standards set by NFPA 1225, 40-50 % of the time
  • 80% or more are from wireless devices

​

3 of 12

DEI of 911 Emergency call

  • PSAP provides services in English. Some centers include Spanish
  • They do have third-party translation service but delays still occur and being put on hold is common
  • Translation back and forth between the caller, translator, and call-taker leads to additional time wasted
  • Cities such as Chicago is home to 153 different languages with Spanish, Polish, Arabic, Tagalog, and Chinese being the most common
    • 36% of Chicago residents speak a language other than English at home

​

4 of 12

PROPOSED SOLUTION

An app that lets the caller converse in their comfortable language. The app is aimed to let them:

    • Text in case they are not in a situation to talk
    • Speak in the language comfortable to them
    • Translate dispatcher’s response to caller language

EmAA is also equipped to:

    • Provide suggested first response for caller accordance to the situation
    • Collect information from caller about the situation while waiting to be connected
    • Summarize the info from caller to dispatcher for faster response

​

5 of 12

FLOW DIAGRAM OF EmAA

User inputs: Voice/ Text

Language Translation

LLM gpt2-large

Dispatcher Summary

    • Location
    • Incident
    • Severity
    • Hazards, etc

Suggestions for the caller

6 of 12

SPEECH RECOGNITION AND TRANSLATION

Speech to Text conversion (online/offline) using Google Speech Recognition

Language detection from text using Google Translate

Converting detected text to English text using Google Translate

Feeding English text to LLM agent

Converting LLM agent English output text to detected language text using Google Translate

Converting output text to speech using Google Text to Speech

7 of 12

GEN AI

Using Gen AI (Large Language Model - gpt2-large) to assist 911 dispatchers helps us in:

  • Improving Emergency Response
  • Enhancing Situational Awareness
  • Supporting Multilingual Communication:s
  • Reducing Response Time
  • Mitigating Human Error
  • Increasing Accessibility

8 of 12

USER INTERFACE

  • When in emergency user opens the app; it asks how they would like to converse
                  • They could choose to speak or text
                  • Don’t have to call 911, app connects them

9 of 12

USER INTERFACE

  • If they choose to speak, EmAA asks them to provide a description of the situation
  • If they choose to text a chatbox appears waiting for their input
  • Translates their input to English while they wait to be connected

10 of 12

USER INTERFACE

  • EmAA then outputs Dispatcher Summary (Sends it away to dispatcher) and suggestion for the caller while they wait

11 of 12

Demo

12 of 12

Future Development

  • Integrate it with PSAP and 911 systems
  • Analyze severity from caller input and assign priority based on 911 protocol
    • Could be used to connect higher severity caller first
  • Include multi languages real time translation
  • Cross question caller to obtain as much information as possible
  • Train LLM to provide better suggestions to the caller