A Multi-Dictionary Approach to Abstractness/Concreteness-Based Authorship Attribution


 We present some early results from a research project aimed at exploring the usefulness of abstractness/concreteness as stylistic features for authorship attribution. conjecture that authors use abstract/concrete words and phrases in suf- ficiently unique ways, so machine learning classifiers can learn to distinguish individual authors’ writing styles. Our approach is based on using abstractness rat- ings texts with established au- thorship generate training vectors different classifiers. The combined word/phrase ratings are extracted two separate dictionaries – an yields stronger than single ab- stractness dictionaries. paper describes details our methodology compares those obtained traditional attribution features. limitations current directions further outlined end paper.