Abstract
We disclose a method and system for measuring genuine structure in a small corpus of short symbol strings (e.g. an undeciphered sign system of a few thousand short inscriptions) in a way that is invariant to any semantic interpretation of the symbols and that requires every claimed pattern to beat a statistical null model preserving the corpus's real nuisance structure. The method comprises: (1) a two-part minimum-description-length "honest score" provably invariant under bijective relabeling of the symbol set, so that proposing a "reading" cannot change the score; (2) a graduated family of maximum-entropy null generators (a "null ladder") that hold fixed, at each rung, string lengths, order-(k−1) symbol statistics, and damage/lacuna positions, generated by count-preserving edge swaps; and (3) an admission gate that credits a pattern's contribution to the score only if it clears the appropriate null rung, awarding zero credit to any proposed reading. This document is published as a defensive disclosure to place the method and its combination in the public domain as prior art. The authors assert no patent rights over the disclosed matter.
Creative Commons License
This work is licensed under a Creative Commons Attribution 4.0 License.
Recommended Citation
Karde, Pravin Dhanraj, "A Reading-Proof Honest-Scoring Method and Nuisance-Preserving Null Ladder for Analyzing Small, Unlabeled, Damaged Symbol Corpora", Technical Disclosure Commons, ()
https://www.tdcommons.org/dpubs_series/11711