SAR: A Graph-Based System with Text Stream Burst Detection and Visualization

Abstract

Smart city trend with Artificial Intelligence, Internet Of Thing and Data Science has been attracting a lot of attention. Following this trend, smart applications that help users improve their quality of life, as well as work, has been investigating by many researchers. In an era of industry 4.0, collecting and exploiting information automatically is essential so that many studies have proposed models for solving storage problems and supporting efficient data processing. In this paper, we introduce our proposed graph-based system called SAR (Smart Article Reader) that can store, analyze, exploit and visualize text streams. This system first gathers daily articles automatically from online journals. After articles are collected, keywords’ frequency of existence is calculated to rank the importance of keywords, finding worthy topics and visually display the results from user requests. Especially, we present the application of Burst Detection technique for detecting periods of time in which some keywords are unusually popular. This technique is used for finding trends from online journals. In addition, we present our method for rating keywords, which share similar Bursts patterns, based on their term frequencies. We also perform system algorithm testing and evaluation to show its performance and estimate its responding time.