New metrics offer interpretable approaches to authorship attribution

Researchers have introduced two novel metrics, Rank-Turbulence Delta and Jensen-Shannon Delta, to enhance authorship attribution by generalizing classical Delta methods. These new measures utilize distance functions designed for probabilistic distributions, offering a more interpretable approach to analyzing word-frequency vectors. The effectiveness of these metrics was demonstrated across literary corpora in English, German, French, and Russian, with Jensen-Shannon Delta consistently matching or surpassing the performance of established methods. AI

IMPACT Introduces new quantitative methods for stylometric analysis, potentially aiding in the interpretation of text generation models.

RANK_REASON Academic paper introducing new metrics for authorship attribution.

Read on arXiv cs.CL →

• Burrows's Delta
• Dmitry Pronin
• Jensen-Shannon Delta
• Project Gutenberg
• Rank-Turbulence Delta
• SOCIOLIT corpus

• paper
• other

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →