Automated Multidimensional Analysis of Global Events With Entity Detection, Sentiment Analysis and Anomaly Detection

Abstract

The modern era, with information overload, has compelled strategic decision makers to obtain assistance from artificial intelligence- (AI) based decision support systems (DSS). Data-driven DSSs powered by powerful AI algorithms can instantly categorize unstructured data, assign context, and produce meaningful insights from such data. This paper presents an automated media monitoring system that can analyze unstructured global events reported in online news, government websites, and major social media to produce significant insights with explainable AI. Using this innovative system, a decision maker can focus on global events requiring urgent attentions since it segregates millions of unnecessary data by using the presented methodology involving entity detection, sentiment analysis, and anomaly detection. The system was designed, deployed, and tested during June 2, 2021 – September 1, 2021. During this 92-day period, the system connected to 2,397 distinct types of news sources and automatically fetched 22,425 major event descriptions from 192 countries. Then, to assign meaning and context to the unstructured event descriptions, the system performed AI-based entity detection and sentiment analysis of these global events. The proposed system is sufficiently robust to detect anomalies instantly from $2.76\times {10}^{8404}$ possible scenarios and provides detailed explanations by using natural language descriptions along with dynamic line charts and bar charts to portray detailed reasoning. The entity detection algorithm had a F1-score of 0.994 and the anomaly detection algorithm had an area under curve score of 0.941, establishing the proposed system with explainable AI as the most accurate, robust media monitoring system according to the literature.

Keywords

• Anomaly detection
• decision support system
• entity detection
• media monitoring with explainable AI
• sentiment analysis