Determination of the defining features of texts written in isolation with a Naive Bayesian Classifier

This study seeks to identify differences between textual samples written in isolation and controls. Isolation is the state of deprivation of one's typical level of social interaction and falls into three categories: prison, seclusion, and isolation. We coded a Naive Bayesian Classifier using the Python package NLTK and ran it with different training to test set ratios and a Leave One Out with authors. The results yielded that accuracy is proportional to training set size. Currently we are analyzing the key features the classifier used to sort the texts and calculating a chance value for the classifier. This is a highly relevant area of study because we hope to elucidate key differences in the thoughts and cognitive states of isolated people, which could predict behavior for socially isolated people.