+- The Voynich Ninja (https://www.voynich.ninja)
+-- Forum: Voynich Research (https://www.voynich.ninja/forum-27.html)
+--- Forum: Analysis of the text (https://www.voynich.ninja/forum-41.html)
+--- Thread: Tests for Simple Solutions (/thread-2706.html)
Tests for Simple Solutions - farmerjohn - 27-03-2019
Here is a try to make list of substantial tests which may be useful to evaluate proposed simple solutions.
Distilled, extreme descriptions are given. Pros and cons of applicability are considered.
It's assumed, that solution consists of the Algorithm which converts Voynichese to Text and claims that it is written in Language. The Algorithm must also have ability to convert Text and Language back to Voynichese.
Letter frequency test
Letter frequencies of Text must match ones of Language.
There statistical methods to compare vectors of frequencies are needed.
+ It may be the only test which not only tests, but is also able to find solution.
+ It may cope with "null symbols" by considering relative letter frequencies.
+ Allows a lot of variations - frequency of adjacent, first, last letters, etc.
Word frequency test
Word frequencies of Text must match ones of Language.
There statistical methods to compare vectors of frequencies are needed.
+ May be also useful if Voynichese word for some reason matches several Language words.
- Certain words and word forms may be ommitted due to stylistical reasons, which affects results in unknown way, so comparison methods there may be much trickier.
Word combination test
Translations of the most frequent Voynichese word combinations are meaningful and common.
+ Good test for repetitive text and when at least something is known about sentence structure.
- More subjective than others.
You are not allowed to view links. Register or Login to view.
Algorithm must distinguish meaningful and meaningless Voynichese text.
Number of test cases depends on required precision, but it should be quite high (6 originally proposed by the author are obviously not enough).
+ Ideal test for "black boxes".
- Number of false positives is unknown in advance and may be quite high.
- In some sense manuscript itself is big "Klaus' test", so for sufficiently advanced Algorithm this test is redunant.
Rene's test
Text written in Language when transformed by the Algorithm to Voynichese must possess all properties of Voynichese.
Need method to determine if given text is Voynichese or not. It may not match Rene's proposal exactly, since the test requires not only convertability (which goes without saying) but also matching certain criteria.
+ Ideal test to remove pair <Algorithm, Language> from consideration.
- "Null symbols" can affect results in non-obvious way.
- Difficult to perform when Language is constructed language - no source texts.
Productivity test
Each attested Voynichese word can be constructed with the Algorithm.
It's may be somewhat complicated to perform this test manually.
+ Analysis of words which do not pass this test may point to erroneous part of the Algorithm.
- Proper names, composite words, "null symbols" are substantial obstacles.