Data Mining the Rabbinic Citation Network

The Talmud Bavli is a semi structured document. The redactors followed certain conventions in compiling the various tractates, including using a common notation in describing trident chains. Thus, patterns like “amar Rabbi A amar Rabbi B” and “Rabbi A amar mishum Rabbi B” are used to identify citations of Rabbi B by Rabbi A. This characteristic can be used to computationally compile a list of citations in the Bavli. This list defines a Rabbinic Citation Network which is amenable to the particular type of data mining known as social network analysis.

This paper mines the Rabbinic Citation Network in 2 different analyses. The first analysis computes a “tradent strength” value between all pairs of amoraim. These values can be visualized in a heat map in order to understand how these strengths correspond to our intuitive knowledge of the amoraic relationships. We can also proceed mathematically, as each amora can now be characterized mathematically by a vector consisting of his “tradent strength” values with all other amoraim. We can use these vectors to calculate the distance between each pair of amoraim and visualize the distances with a heat map or cluster the amoraim using conventional clustering algorithms such as k-means. These techniques separate the amoraim into different classes that help us understand the different paths along which tradition was transmitted through the amoraic generations.

The second analysis utilizes the Rabbinic Citation Network to perform an ab initio computation of the amoraic generations, which can then be compared to the conventional assignment of generations to the amoraim. We consider how these changed values may change our understanding of the amoraic transmission process.