An Artificial Intelligence-Driven Mobile Application for Real-Time Assistance in Taxi-Related Criminal Incidents Using Natural Language Processing in Metropolitan Lima

Abstract

Herein, SmartSecurity, an artificial intelligence (AI)-based mobile application designed to enhance passenger safety in taxi services across Metropolitan Lima, is proposed herein. This system detected predefined emergency keywords via real-time voice processing and automatically activated geolocation and multichannel alerts via WhatsApp and SMS. Methodologically, an applied experimental–developmental design was adopted. The application was implemented using Flutter for the front-end interface and FastAPI on Google Cloud Platform for backend services. The Whisper AI model specialized in multilingual speech recognition was fine-tuned using contextualized Spanish audio datasets. The model performance was evaluated across parameters such as accuracy, sensitivity, F1-score, false-positive rate, and latency under simulated taxi conditions. The application achieved 90% accuracy and sensitivity, an F1-score of 0.90, and a near-zero false-positive rate, with an average latency of 1 s and a total response time of 2 s. The area under the ROC curve (AUC) approached 1.00, indicating high discriminative capacity of the model even in noisy environments. The proposed model thus provides an efficient, low-cost, and robust solution to the rising issue of urban transport insecurity, contributing to safer and smarter mobility systems in Metropolitan Lima.