The Voynich Ninja

Excuse me, Geoffrey, I didn't follow in detail your proposal to read the glyphs in the manuscript, on what do you base the attribution of phonetic values, did you do a frequency analysis, or do you rely on your intuition?

(25-09-2020, 12:34 PM)Ruby Novacna Wrote: You are not allowed to view links. Register or Login to view.Excuse me, Geoffrey, I didn't follow in detail your proposal to read the glyphs in the manuscript, on what do you base the attribution of phonetic values, did you do a frequency analysis, or do you rely on your intuition?

Most of the details are discussed in the long "Verbose cipher" thread, but I can try to provide a short summary here:

The most significant new result that I observed in the best version (highest conditional entropy) of Koen's and Marco's verbose cipher analysis of the Voynich ms text was the treatment of EVA [o+glyph] as a single unit. Given the relatively high frequency of both [k] and [ok], [t] and [ot], etc., it makes logical linguistic sense to hypothesize that [o+glyph] may represent the voiced counterpart of voiceless [glyph] for obstruent phonemes (such as /p/ & /b/, /t/ & /d/, /k/ & /g/, /s/ & /z/, etc.). This hypothesis is based on typical frequency patterns of voiceless and voiced obstruents in most natural languages. 

Then I considered the glyph(s) EVA [ch] and its combinations and ligatures with other glyphs that are well-known, [ckh] and [cth] being the most significant of them. Now this could logically represent a variety of linguistic features: for example, [ch] could represent /h/ and its combinations and ligatures could represent aspirated counterparts of the phonemes it combines with. However, I observed that [ockh] and [octh] occur as a proportion of all [ckh] and [cth] much more frequently than does [och] as a proportion of all [ch]. If [ch] were aspiration, then the combination of voicing and aspiration in [ockh] and [octh] would be unexpected in most natural languages--yes, it famously occurs in South Asian languages derived from Sanskrit, in the sounds "bh", "dh", "gh", but closer to the likely area of origin of the Voynich ms such sounds are very rare. So I rejected the hypothesis of [ch] as /h/ and aspiration.

I found it makes much more logical linguistic sense to hypothesize that [ch] represents /i~j/ and its combinations and ligatures represent palatalization. The combination of voicing and palatalization in [ockh] and [octh] would be perfectly natural in languages with palatalized consonants, such as the Slavic languages. Furthermore, the relative rarity of [och] would also make sense, as there is no such thing as a voiced/voiceless distinction in the palatal glide /j/. 

Remarkably, I have recently discovered that You are not allowed to view links. Register or Login to view. a couple months ago independently suggested the same hypothesis of [ch] as /j/ and ligatures with [ch] as palatalized consonants in Slavic languages! The post even concluded by suggesting "Sorbian or Judaeo-Czech" as plausible languages of the Voynich ms! 

Another important idea in Koen's and Marco's work was the treatment of EVA [qok] and [qot] as single units. Since my hypothesis already treats [k] glyphs and combinations as dental/alveolar phonemes and [t] glyphs and combinations as labial phonemes, I had the idea to hypothesize that perhaps [qok] = /n/ and [qot] = /m/. 

Those are the most important broad hypotheses behind my current interpretation, which could perhaps be called "Slavic VCI". (I suppose "Selective Letter Alphabet for the Voynich Inventory of Characters - Verbose Cipher Inventory" would be too much.) There are also logical linguistic reasons for other details of my hypothesis: for example, the great frequency of the combination EVA [lk] led me to hypothesize that [l] = /s/ and [k] = /t/, with various sequences with [l] representing various sibilant consonants. Once one's hypothesis is focused on Slavic, then EVA [r] = /l/ makes logical linguistic sense due to the frequency of final /l/ in Slavic past tense verb forms. It remains to assign EVA [s] as /r/. 

EVA [d] and [od] fit in the logical linguistic system best as /k/ and /g/, but my hypothesis is that variations in the shape of EVA [d] may distinguish /c/ from /k/ in many cases. Also, I hypothesize that [od] would represent /h/ rather than /g/ in a West Slavic language, in line with the actual historical sound change /g/ > /h/. An earlier version of my hypothesis had treated [o] in isolation as /h/, but now I interpret it as Slavic "ch", the voiceless velar fricative /x/. 

So, is that frequency analysis, or is it intuition? I would describe it as intuition based on typical frequencies of phonemes and combinations of phonemic features in the linguistic area. 

Geoffrey

(25-09-2020, 04:02 PM)geoffreycaveney Wrote: You are not allowed to view links. Register or Login to view.is that frequency analysis, or is it intuition? I would describe it as intuition based on typical frequencies of phonemes

Thanks Geoffrey, I like the short and clear summaries.

(25-09-2020, 06:27 PM)Ruby Novacna Wrote: You are not allowed to view links. Register or Login to view.Thanks Geoffrey, I like the short and clear summaries.

You're welcome, Ruby. Also, please click on the "Add Thank You" button on my posts as well, if you are thankful and if you don't mind. Some folks may keep track of or care about the "Thanks Received" statistics displayed on this site 

Nulls are often mentioned when it comes to possibilities we can expect, so it is on my "to check" list. However, the verbose cipher thing took me so long that I moved on to imagery for a while. 

What I was planning was to check what would happen for each glyph if it was treated as a null. And combinations of glyphs, like what if [a, o and y] are all nulls? I don't understand why [y] should be a special case though. Take [a] for example; instead of merging it with each possible combination of following glyphs, it may be simpler to just omit it.

(25-09-2020, 07:24 PM)Koen G Wrote: You are not allowed to view links. Register or Login to view.Nulls are often mentioned when it comes to possibilities we can expect, so it is on my "to check" list. However, the verbose cipher thing took me so long that I moved on to imagery for a while. 
What I was planning was to check what would happen for each glyph if it was treated as a null. And combinations of glyphs, like what if [a, o and y] are all nulls? I don't understand why [y] should be a special case though. Take [a] for example; instead of merging it with each possible combination of following glyphs, it may be simpler to just omit it.

Your [o+glyph] bigrams worked so well on the h2 conditional entropy of the text that it's hard to believe the results could be as good or better if [o] were treated as a null instead. And you also achieved your best results with many n-grams including [a] as well. If you just omit [a], you are losing for example the distinction between [al] and [l], and between [ar] and [r]. With [y] on the other hand I don't recall any n-grams with [y] that ended up in your best results to improve the entropy statistics. And there is also the linguistic argument that the ubiquitous vord-final [y] excessively reduces the number of possible word-final or syllable-final letters/phonemes. Treating [y] as a null really opens up many more actual meaningful vord-final letters representing actual phonemes. I think it makes it much easier to relate to actual natural languages and texts.

Hmm yes, it might work like that, it's worth a shot. I should have some time tomorrow to play around with this.

It's not always easy to predict though. For example, why is merging [ch] recommended for entropy optimization but merging [dy] isn't? Nor is [ed], for example. That's why I was planning to test all characters as nulls. It would be interesting if one glyph or a class of glyphs stands out.

(25-09-2020, 07:24 PM)Koen G Wrote: You are not allowed to view links. Register or Login to view.What I was planning was to check what would happen for each glyph if it was treated as a null.

As I recall, the most likely null I found when I went looking for them was line-initial EVA s-.

(25-09-2020, 08:00 PM)Koen G Wrote: You are not allowed to view links. Register or Login to view.Hmm yes, it might work like that, it's worth a shot. I should have some time tomorrow to play around with this.
It's not always easy to predict though. For example, why is merging [ch] recommended for entropy optimization but merging [dy] isn't? Nor is [ed], for example. That's why I was planning to test all characters as nulls. It would be interesting if one glyph or a class of glyphs stands out.

Thanks Koen. Let us know the results, I'm looking forward to seeing them! 

As for [ch] vs. [dy] or [ed], sure [d] occurs as part of [dy] in over half of its occurrences, and as part of [ed] in almost half of its occurrences, but there are still thousands of occurrences of [d] without following [y] or preceding [e] -- such as the numerous occurrences of [daiin], et al., for example. 

With [ch], on the other hand, you can almost never find [c] in the ms text without [h] either immediately following or as part of a ligature [c+glyph+h]. So [ch] is rather more clearly a single unit.

(25-09-2020, 06:51 PM)geoffreycaveney Wrote: You are not allowed to view links. Register or Login to view.

(25-09-2020, 06:27 PM)Ruby Novacna Wrote: You are not allowed to view links. Register or Login to view.
Thanks Geoffrey, I like the short and clear summaries.

You're welcome, Ruby. Also, please click on the "Add Thank You" button on my posts as well, if you are thankful and if you don't mind. Some folks may keep track of or care about the "Thanks Received" statistics displayed on this site 

I hope the most important criterion for thank you is the quality of the research together with basic courtesy and appreciation.

For the record, I do not keep track of Thank Yous. I think of them like sunshine. You shouldn't be obsessing over when it will come out. Simply appreciate it when it happens.