You are not allowed to view links. Register or Login to view. include several works related to the issues of transliteration, word morphology and higher structure (e.g. word order / line grammar).
Rene discusses new progress in processable transliterations. Alexander Boxer will discuss his new transliteration and some aspects of word order (e.g. token repetition).
Massimiliano Zattera's paper is about both transliteration and morphology.
I like the idea of applying the You are not allowed to view links. Register or Login to view. to this problem. This measure takes into account the ability of the method to both detect correct samples (words from a Voynich transliteration, in this case) and reject incorrect samples (non-Voynichese words). I think a similar approach could usefully be applied to a grammar of Voynichese lines: such grammar should accept actual lines and reject members of a set of non-Voynichese lines.
I am curious to see how Zattera tackled the problem of creating the set of negative samples: likely there are well-established ways to do so in the context of computational linguistics, but this is something I have not investigated. When I did something similar to compare Stolfi's and Emma's word-grammars, I scrambled characters in Voynichese words and created a set of the scrambled results that do not occur in the manuscript; I also included a set of English words filtering out the few that occur in EVA (e.g. 'do', 'chain') and those that contain characters that are absent or very rare in EVA (e.g. 'b', 'w'). When investigating line-grammar, negative cases could be created by swapping words in a Voynichese line.
I would say that the majority of the other papers at Malta are more or less related with aspects of Voynichese "grammar". I am sure that we will have much to digest and discuss after the conference...
Rene discusses new progress in processable transliterations. Alexander Boxer will discuss his new transliteration and some aspects of word order (e.g. token repetition).
Massimiliano Zattera's paper is about both transliteration and morphology.
Quote:The question of which glyphs are actual single Voynich characters is still very open; its correct answer will greatly impact statistical analysis of the text and will guide deciphering attempts. This research uses a computational approach, including machine learning techniques, to create a new transliteration alphabet.
...
This alphabet has been used by an optimization algorithm to automatically produce a formal grammar for words in the Voynich. This grammar is the one with the highest F1 score among those surveyed.
...
I like the idea of applying the You are not allowed to view links. Register or Login to view. to this problem. This measure takes into account the ability of the method to both detect correct samples (words from a Voynich transliteration, in this case) and reject incorrect samples (non-Voynichese words). I think a similar approach could usefully be applied to a grammar of Voynichese lines: such grammar should accept actual lines and reject members of a set of non-Voynichese lines.
I am curious to see how Zattera tackled the problem of creating the set of negative samples: likely there are well-established ways to do so in the context of computational linguistics, but this is something I have not investigated. When I did something similar to compare Stolfi's and Emma's word-grammars, I scrambled characters in Voynichese words and created a set of the scrambled results that do not occur in the manuscript; I also included a set of English words filtering out the few that occur in EVA (e.g. 'do', 'chain') and those that contain characters that are absent or very rare in EVA (e.g. 'b', 'w'). When investigating line-grammar, negative cases could be created by swapping words in a Voynichese line.
I would say that the majority of the other papers at Malta are more or less related with aspects of Voynichese "grammar". I am sure that we will have much to digest and discuss after the conference...