Higher Criticism for Discriminating Word-Frequency Tables and Testing Authorship

TL;DR

This paper introduces an adaptation of the Higher Criticism test to measure similarity between word-frequency tables, improving authorship attribution accuracy and identifying key discriminating words with low variance.

Contribution

It presents a simple, tuning-free method that enhances authorship attribution by effectively measuring document similarity and highlighting characteristic words.

Findings

Achieves state-of-the-art accuracy in authorship attribution challenges.

Identifies low-variance, author-specific discriminating words.

HC-based measure is robust to topic variations.

Abstract

We adapt the Higher Criticism (HC) goodness-of-fit test to measure the closeness between word-frequency tables. We apply this measure to authorship attribution challenges, where the goal is to identify the author of a document using other documents whose authorship is known. The method is simple yet performs well without handcrafting and tuning; reporting accuracy at the state of the art level in various current challenges. As an inherent side effect, the HC calculation identifies a subset of discriminating words. In practice, the identified words have low variance across documents belonging to a corpus of homogeneous authorship. We conclude that in comparing the similarity of a new document and a corpus of a single author, HC is mostly affected by words characteristic of the author and is relatively unaffected by topic structure.