The tannaitic legal Midrashim did not all survive and are not all known to us in a complete independent form. David Zvi Hoffman was one of the first scholars to recognize the 13th century Yemenite Midrash, Midrash haGadol, written by R. David of Aden, as a major source of the lost legal Midarshim. He published the Midrash Tannaim, containing all of the tannaitic looking paragraphs from Midrash haGadol on the book of Deuteronomy. However, the author of Midrash haGadol often introduced changes into the material he borrowed from rabbinic and medieval sources. The resulting passages often seem to be unparalleled tannaitic sources, when in fact they are not. This article proposes a re-examination of the Mekhilta material as found in the Midrash haGadol, in order to reconstruct more accurately the tannaitic text. We propose a methodology for contending with this challenge, via a new approximate-matching algorithm designed to identify modified sources of this sort. Using this algorithm, we first compared Hoffman’s Midrash Tannaim on Deuteronomy to the Sifre, filtering out all parts of the text that are simply reworkings of the Sifre, despite many interpolations, omissions, and modified words. Having removed the Sifre passages from within the Midrash Tannaim text, we then proceeded to the next stage, in which we investigated the presence of reworked Maimonidean excerpts within the remaining text. The Maimonidean excerpts pose a particular challenge, because their reuse in the Midrash haGadol involves not only modifications and interpolations, but also changes of order. We describe the modifications that were necessary to the algorithm in order to handle these out-of-order cases of reuse as well. We have thus far succeeded in identifying and removing the reworked material appropriated from the Sifre and from Maimonides, and in the future we plan to tweak the algorithm such that it can successfully identify additional rabbinic passages as well, including the Babylonian and Palestinian Talmudic material, and other midrashic compilations. This will ultimately allow us to produce a final text approximating the original Mekhilta, to the greatest extent possible.
Performance Testing of GPU-Based Approximate Matching Algorithm on Network Traffic