Estimation of Community Views on Criminal Justice a Statistical Document Analysis Approach
Sujeong Seo *
School of Mathematical Sciences, Rochester Institute of Technology, Rochester, NY 14623, USA.
Ernest Fokoue
Faculty of School of Mathematical Sciences, College of Science, Rochester Institute of Technology, Rochester, NY 14623, USA.
*Author to whom correspondence should be addressed.
Abstract
The Community Views on Criminal Justice System (CVCJS) initiative was established to collect a city community's perceptions on experiences with local Police Departments and other agencies in the criminal justice system, and share those findings to inform local Gun Involved Violence Elimination (GIVE) strategies in New York State. This paper reviews those findings via an empirical study with major text mining methods. Specifically, atomic/canonical words along with as n-grams are used to explore such text mining tasks as sentiment analysis, document clustering and topic modeling, all aimed at gaining insights into all the patterns underlying the community's perception of policing and criminal justice. We use Latent Dirichlet Allocation [LDA] analysis and Structural Topic Model [STM] analysis, which are currently among the most widely used topic modelling algorithms in the fields of computer science, statistics, and machine learning. Despite the very limited amount of data available for our study, the combination of sentiment analysis with document clustering and topic modelling helps extract and reveal very interesting patterns underlying the community's views of policing and criminal justice.
Keywords: Text mining, document clustering, sentiment analysis, topic modeling, n-gram, criminal justice, data science, statistical analysis.