Automatic classification of documents by formality

This paper addresses the task of classifying documents into formal or informal style. We studied the main characteristics of each style in order to choose features that allowed us to train classifiers that can distinguish between the two styles. We built our data set by collecting documents for both styles, from different sources. We tested several classification algorithms, namely Decision Trees, Naïve Bayes, and Support Vector Machines, to choose the classifier that leads to the best classification results. We performed attribute selection in order to determine the contribution of each feature to our model.

1. Introduction

The need for identifying and interpreting possible differences in linguistic style of texts, such as between formal and informal styles, has increased nowadays as more and more people are using the Internet as a main resource for their researches. There are different factors that affect formality, such as words and expressions, as well as syntactical features. Vocabulary choice is perhaps the biggest style marker. Generally speaking, longer words and Latin origin verbs are formal, while phrasal verbs and idioms are informal. There are also many formal/informal style equivalents that can be used in writing.