Real-Time Scam Detection and Sentiment Analysis: An NLP-based Multilingual Approach

As communication faced a digital transition, the number of scam calls increased tremendously, leading to severe threats to security and finances. Most of the traditional detection systems fail in real-time deception analysis and are extremely rigid to changes in the scam strategies. In this paper, we propose an enhanced audio conversations analysis for scams detection system based on NLP and audio signal processing. The system proposed in this paper captures and analyses audio streams containing conversation dynamically to extract fraud intent linguistic and acoustic features. The use of advanced deep speech recognition models for analysing speech together with natural language processing (NLP) based intended action identification increases efficiency and precision of the system. The system was developed to be easily integrated into existing telecommunication networks and automated customer service systems to reduce fraud opportunities with no or minimal delay.