Classification of calls received by the emergency call center using natural language processing methods: Bilecik 112 emergency call center example

Emergency call centers are among the critical infrastructures that are vital for the safety of life and sustainability of social order in modern societies. This study proposes an innovative automatic classification system based on natural language processing (NLP) in order to increase operational efficiency and optimize response times in the processing of emergency calls. Within the scope of the research, voice recordings obtained from a real call center environment were first converted into textual data and then subjected to a comprehensive data preprocessing process. In this process, basic NLP techniques such as text normalization, tokenization, stop-word elimination and anonymization of proper nouns were applied. As a machine learning approach, Naive Bayes, Support Vector Machines (SVM), Random Forest and Logistic Regression algorithms as well as deep learning-based LSTM, GRU,CNN and BERT models were comparatively evaluated. Model performances were measured with precision, sensitivity, and F1-score standard accuracy metrics. Experimental results showed that different accuracy results were achieved in the developed models. These findings reveal that artificial intelligence-based solutions in emergency management systems can significantly increase operational efficiency and minimize human-related errors. It is anticipated that the study will contribute to the digital transformation efforts of emergency services in the future.