Andrews University Seminary Studies, Vol. 59, No. 2, 251-271.
Copyright © 2022 Andrews University Seminary Studies
APPLICATION OF THE TERM FREQUENCY-INVERSE
DOCUMENT FREQUENCY WEIGHTING SCHEME
TO THE PAULINE CORPUS
BranpDoN [.S. vAN DER VENTEL RicHArD T. NEWMAN
Stellenbosch University Stellenbosch University
Abstract
The term frequency-inverse document frequency (TF-1DF) weight-
ing scheme is applied to the text of the thirteen epistles tradition-
ally associated with the apostle Paul. The data for the analysis is
the morphologically tagged text of the Society for Biblical Litera-
ture’s Greek New Testament. The TF-IDF scheme is then used to
construct the document term matrix (DTM) for a corpus under
consideration. The DTM allows each document to be represented
by a multi-dimensional document vector. A query document is
then chosen and a vector representation of it is constructed. The
cosine similarity between the query document and documents
in the corpus is calculated. The following pairs of documents are
consistently found to have the highest similarity: (1) Romans and
Galatians, (2) Ephesians and Colossians, and (3) 1 Timothy and
Titus. It is shown that computational methods may be applied to
the thirteen epistles and that the results are in accordance with those
obtained from theological or literary analysis.
Keywords: New Testament, Paul, authorship, term frequency-inverse
document frequency
Introduction
The New Testament was written in Koine Greek, an ordinary human language,
which implies that the text can be analyzed using methods of computational
linguistics.! Stylometry is one aspect of computational linguistics in which
a statistical analysis of a corpus is done. One application of stylometry is
authorship-attribution and one of the earliest examples, applicable to biblical
studies, is that of Augustus De Morgan who, in 1851, suggested that average
word length could be used for author-attribution. In a letter addressed to a
' William D. Mounce, Basics of Biblical Greek (Grand Rapids, MI: Zondervan,
2009), loc. 457.
251