Keyword and named entity recognition on emergency call hotline data

Abstract

One of the most important services provided by the healthcare industry is emergency medical services. This service is engaged by the use of an emergency call hotline. Important information is being taken note of by the medical professional operating the hotline from the caller. Based on the information received, these professionals have to suggest a responsive next course of action which will be crucial based on the severity of an emergency. Therefore, the hotline operators must be able to identify key and necessary information when dealing with the caller. This report will discuss Named Entity Recognition (NER) application on emergency call hotline conversation data such that this system helps the medical professional to identify key information faster and more accurately and improve their response time. A set of emergency sentences will be created based on grammar rules that were extracted from multiple datasets. This set of sentences will be used to train a Bi-LSTM-CRF model to implement a NER system effectively.