Abstract
This essay considers the debate on the diachrony of Biblical Hebrew from the perspective of pattern recognition , guided by the precepts of meta-analysis ( systematic review ). The necessity of taking possibly confounding variables into account is stressed, as is the central importance of three types of noise in the data: transmission noise (changes introduced as the texts were passed along), feature noise (imprecision due to textual ambiguity), and class noise (contamination of one corpus by another). On the one hand, the nature of the corpus-discriminating features proposed in the literature suggests that the enterprise has been severely compromised by overfitting , wherein too many locally-applicable features have been used to characterize limited texts and then have been inappropriately generalized across Early Biblical Hebrew and Late Biblical Hebrew. On the other hand, when Young, Rezetko, and Ehrensväld rely on 500-word samples in assessing the literature, they err as these samples are too small. Further, their definition of accumulation is deficient as it weights features equally and takes little account of noise. Pattern recognition theory provides superior ways both of identifying and combining evidence.
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have