Authorship verification of the Deutero-Pauline letters through deep learning
Evy Beijen and Rianne de Heide
Authorship verification of the Deutero-Pauline letters through deep learning
Evy Beijen and Rianne de Heide
HIPHIL Novum, 10(1), 22–39, 2025. https://doi.org/10.7146/hn.v10i1.147482 proc
What is this paper about?
This interdisciplinary paper uses deep learning for a classical authorship question: which disputed New Testament letters look statistically like the undisputed Pauline letters? It trains bidirectional LSTM models on Greek text and compares their classifications across disputed letters.
Summary
In the Christian tradition, fourteen letters of the New Testament have been attributed to the Apostle Paul, while the authorship of seven of them has been debated. This study develops bidirectional long short-term memory networks to classify approximately 100-word chunks of disputed letters as Pauline or non-Pauline. Plaintext and lemmatized variants are trained on undisputed Pauline letters and negative-example “impostor letters”. The two variants achieve similar accuracies and agree most clearly on Colossians and 2 Thessalonians as more Pauline, and Hebrews and 1 Timothy as more non-Pauline, while giving less stable classifications for Titus, 2 Timothy and Ephesians. The results are interpreted cautiously, and the study is presented as a deep-learning approach that may support future work on authorship verification of the Pauline corpus.