One of the world’s top computing teams for authenticating authorship can be found at Duquesne University.
A team working in Duquesne’s Evaluating Variations in Language (EVL) lab correctly analyzed more documents than any other of the 25 international competitors in the Plagiarism Action Network (PAN), producing of one of the highest overall accuracy records.
Participants were asked to analyze a number of documents to determine the author of each document, including in some cases a “none of the above” choice when they were given a document but no other writings to compare to or learn from by that specific author. They also were asked to analyze some other documents to determine where and how they had been changed, for example, by cutting and pasting new material from another source or by combining paragraphs from several different authors.
The documents came from online sources such as Feedbooks, an e-publishing house. The competition was divided into eight subtasks, and the results will be discussed at the Sept. 17-20 PAN conference in Rome.
Team members DU alumni John Noecker Jr. and Michael Ryan are engineers in Dr. Patrick Juola’s EVL Lab, which is funded in part through the National Science Foundation and the National Endowment of the Humanities, a seldom-seen combination. Through additional grant funding, the team will examine writings that are anonymous or use a pen name, but believed to be written by a young Abraham Lincoln.
“This is an exciting example of cutting-edge science in an important new field,” said Juola, an international expert in authorship authentication and plagiarism detection. “It’s great that Duquesne has been invited to participate in setting it up. And the results speak for themselves. It’s doubly great that the EVL Lab’s work can outscore the rest of the world. John and Mike are brilliant researchers and I feel lucky that they are working here instead of Silicon Valley.”
Juola and his team further are refining the program they have developed that encompasses word usage and writing style to determine the author of writings, from policy documents to suicide notes.