Age Detection in Chat

Abstract

This paper presents the results of using statistical analysis and automatic text categorization to identify an author’s age group based on the author's online chat posts. A Naive Bayesian Classifier and Support Vector Machine (SVM) model were used. The SVM model experiments generated an f-score measurement of 0.996 on test data distinguishing teens from adults. We also introduce an alternative method for generating “stop words” that chooses n-grams based on their relative distribution across the classes.