Stylistic text classification using functional lexical features

Most text analysis and retrieval work to date has focused on determining the topic of a text, what it is about. However, a text also contains much useful information in its style, or how it is written. This includes information about its author, its purpose, feelings it is meant to evoke, and more. This paper addresses the problem of classifying texts by style (along several different dimensions), developing a new type of lexical feature based on taxonomies of various semantic functions of different lexical items (words or phrases). We show the usefulness of such features for text classification by author, author personality, gender of literary characters, sentiment (positive/negative feeling), and scientific rhetorical styles. We further show how the use of such functional features aids in gaining insight about stylistic differences between texts. ∗Casey Whitelaw was a visiting scholar at the IIT Linguistic Cognition Laboratory during November 2004.