George Lasry, a computer scientist and an expert in cryptanalysis, provided me with his "interim results" for an unsolved ciphertext found in BnF fr.4715, f.85. He has solved many historical ciphers (see "Unsolved Historical Ciphers") besides many other achievements outside my scope (Google Scholar). His "interim results" this time are interesting for me in that he calls it a "machine reconstruction."
He developed many related tools for cryptanalysis of historical ciphers, as well as language models based on historical text for algorithms to work effectively (e.g., his paper on papal ciphers: Lasry et al. (2021)). In addition to these, transcription of the ciphertext is the indispensable first step of any computer processing. He developed a manual tool to mark and classify graphic symbols, of which a glimpse can be made in the provisional decipherment below. (Text data is here.)
In this case, Arabic figures 01-57 are assigned to symbols more or less in the order of frequency. Symbols 100-122 are ones occurring only once. His algorithm works for homophonic substitution of single letters, so he excluded 100-122 from cryptanalysis because they may not correspond to single letters, but prepositions, suffices, or even names. Of course, some of 01-57 may be syllables rather than letters, but as long as a sufficiently large proportion of these symbols represent letters, his algorithm may give meaningful fracments of words here and there.
In the above decipherment, there seem to be fragments of some Italian words: "scriver", "parlando", "contra", "francese", "signor", "..nditione", "ma consient...." There must be many errors in this "machine reconstruction", but hopefully someone versed in Italian may correct the assignment starting from these correspondence.
No comments:
Post a Comment