Naïve Bayes Emotion Classification of Final Statements from Death Row Youths before Execution

Abstract

The behaviours of most Death Row inmates in pre-prison era typically involve neurological insult, developmental histories of trauma, family disruption, and substance abuse. These behaviours lead them to commit crimes that finally land them into death row in prison. Rates of psychological disorder among death row inmates are high, where environments and conditions of confinement appear to aggravate such disorders. The last statements of Death row inmates can be deeply emotional. Several Machine Learning (ML) algorithms can be used to classify these emotions. Although simple, the Naïve Bayes algorithm is a popular algorithm that has proven to be robust in text mining. In this study, the Naïve Bayes and benchmark data available from the Texas Department of Criminal Justice to detect and classify emotions from the lasts statements of youth executed on death row. Friedman’s test with Bonferroni Adjustment was employed to examine whether the executed inmate’s ethnic race has an effect on the emotion of their final statement. Findings from the study indicated that despite ‘anger’ having both highest minimum and maximum values compared to other emotions, it did not achieve the highest mean score and it came second to the emotion ‘joy’. Joy has 0.31 mean score followed by anger with 0.30, sad with 0.26, love with 0.06, fear with 0.05, and surprise with 0.02. Using the Friedman’s test with Bonferroni adjustment, this study discovered that there was no significant difference between the different ethnic races of executed inmates and emotions from their last statements.