A computational analysis of the special Talmudic tractates

The Babylonian Talmud, primarily composed in Jewish Babylonian Aramaic, exhibits various non-standard linguistic features, interspersed throughout it. Medieval rabbis highlighted some tractates, often referred to as the 'special tractates', which possess a more abundant number of occurrences of these features than others. This identification involved mostly classical linguistic and philological methodologies, along with some scholarly intuition. This article proposes identifying special tractates using quantitative methods, namely machine-learning tools. Our algorithm successfully detected a large percentage of non-Babylonian features in all special tractates. A more nuanced examination revealed a concentration of features in specific stories of Tamid, rather than throughout the tractate. It also points to other tractates which are comparatively close to the special tractates, and several tractates whose dialect is more uniform than the average, a phenomenon not previously investigated. We also discuss the advantages and disadvantages of employing machine-learning tools in comparison to human analysis.