N-gram-based Text Attribution

Quantitative authorship attribution refers to the task of identifying the author of a text based on measurable features of the author’s style—a problem that has practical application in areas as diverse as literary scholarship, plagiarism detection, and criminal forensics. Attribution methods generally follow a generative approach, wherein a statistical “profile” is created for a set of candidate authors, based on certain features of the authors’ writings, and the author whose profile most closely resembles the corresponding features of the unclassified text is selected. Potential features include word and sentence lengths, letter frequencies, word frequencies, vocabulary richness, word collocations, and more sophisticated (but not necessarily more useful) patterns that appear after syntactic tagging.