Design of Emergency Call Record Support System Applying Natural Language Processing Techniques

Abstract

Currently Command and Control Centers (C2), such as the Integrated Security Service ECU 911, are managed by Computer Assisted Dispatch Systems (CADS). These systems facilitate the registration of incidents and the distribution of rescue resources. However, the information registration process does not yet have automated methods. Important data such as: name, address, reference and categories are recorded manually, which generates problems of loss of information and inefficiency, in terms of time and attention to the incident. As a solution to these problems, the design of an emergency call record support system is proposed, based on Natural Language Processing (NLP) techniques and algorithms. Taking into consideration an analysis of the processes of the ECU 911, three modules are proposed: (1) transcription of audio to text calls (ASR), (2) extraction of relevant information (NER) such as: address and references; and (3) call classification (TF-IDF/SVM) according to service and priority. Thus obtaining the design of an automated support system for CADS, which provides quality information in a timely manner.