Analyzing Large News Corpus Using Text Mining Techniques for Recognizing High Crime Prone Areas

Abstract

For any developing country, crime rate analysis over different states and different districts for each state is an important task. For having a gross view of crime rates, online newspapers can be used as a useful resource for such tasks. Since all types of crimes are not reported to the police department and the reported crime data may not be easily accessible by the common people, the crime location information can be extracted from online local newspapers and/or national newspapers and the extracted information can be used for counting crime incidents for a location or an area. But, processing the huge collection of newspapers and extracting crime locations from them is not a trivial task. This task can be roughly divided into two main parts, the first part is to filter the crime news articles from the automatically downloaded news collection and the second part is the crime location extraction from each news article. Our proposed system does this task automatically through mining the collection of newspapers written in the Bengali language. To get the picture of high crime prone locations in a state, the proposed system visualizes the crime count per district in a state. Our system is tested on data related to the state of West Bengal. To generate the location wise crime histogram, this system can be used as a tool.