Are perfect-reduplication and quasi-reduplication related?

[Koen G]

Great work, Marco. This seems pretty conclusive - we probably won't find a parallel for Voynichese in your average text.

There are recurring indications that our best bet is to look at poetry/hymns/incantations, but we have not yet found a text with this high degree of reduplication maintained over tens of thousands of words. 

As you say, we are looking for an extreme form of poetic and/or linguistic resonance. Used in an entire manuscript, apparently about various subjects.

The only alternatives are the various scenarios where Voynichese words are not words. Meaningless text, or syllables, or some cypher that generates redundancy.

So I wonder, have we reached the limits of comparing plain texts to EVA transcriptions?

[ReneZ]

So I wonder, have we reached the limits of comparing plain texts to EVA transcriptions?

I think we've only touched the surface.

What has been tried still very little is to experiment with different parsing of the text.

[Koen G]

Rene: yes, that is what I was hinting at. I probably should have said "standard" or "unmodified" EVA transcriptions.

When it comes to different parsing options, I agree we're only scratching the surface and there are more possibilities than we can ever test. It's a line of investigation I find promising and enjoyable though.

[ReneZ]

One thing that seems to have gone lost in time, in a way, is that Eva should not be used for statistical analyses, because it was always clear that the individual Eva symbols represent parts of glyphs.

[Koen G]

For something like Marco's current research, which focuses on vocabulary units and minor variations between them, this is less relevant though. The trickiest part there might be the interpretation of spaces.

[MarcoP]

Hi Koen and Rene,
I am sure that there is much more than can be done. I have been very much focussed on reduplication, but that is just a tiny subset of what can be done. A further step in this same area could be comparing reduplicating words in the VMS and in reduplicating linguistics texts. For instance, among the reduplicating words in the Kalevala, the most common one in the whole text is 'poika' (boy) ranking 21st. In the VMS, the most frequent word 'daiin' appears several times as 'daiin daiin': what to make of this?
The more one compares with other languages, the more differences are found, which is both interesting and frustrating.

On a side possibly more related with parsing, I think experiments with actual transcriptions of handwritten text can be useful. Spelling variations, abbreviations and maybe scribal errors can contribute to generate a number of similar words that actually stand for a single "ideal" word-type. Finding a well-grounded way to "normalize" the VMS by aggregating words on the basis of both their spelling and their behaviour might allow us to find patterns that are obscured by accidental shifts in words. But I don't know if this has any chance of being possible. The task is also related with trying to map (or normalize) the systematic differences between Voynich "dialects".

What kind of approach are you thinking of when you speak of parsing Voynichese in the context of comparison with plain written text?

[Koen G]

Marco: but if the VM uses more patterns more consistently than any known language or similarly sized text so far, what could we learn by normalizing its supposed spelling variations/errors? This would drop large-window MATTR, but that is no problem, it would probably still be well within acceptable levels. But in small-window patterns, it would only widen the gap between Voynichese and other languages, right? 

Or are you rather thinking about patterns of vocabulary related to similar contexts?

[MarcoP]

Marco: but if the VM uses more patterns more consistently than any known language or similarly sized text so far, what could we learn by normalizing its supposed spelling variations/errors?

I think that, even if the manuscript is meaningful, spelling is likely inconsistent, as is the case with other manuscripts. I am not sure that there is anything to learn.

This would drop large-window MATTR, but that is no problem, it would probably still be well within acceptable levels. But in small-window patterns, it would only widen the gap between Voynichese and other languages, right? 

I think so.

Or are you rather thinking about patterns of vocabulary related to similar contexts?

Most identifiable features in the manuscript, like reduplication and quasi-reduplication, Grove words, line-effects, appear problematic. Maybe a lower MATTR would make it possible to find something closer to grammatical patterns. Normalization across dialects could make shared structures detectable. But of course this is all pure speculation.

[ReneZ]

What kind of approach are you thinking of when you speak of parsing Voynichese in the context of comparison with plain written text?

There are quite a number of questions which have not been fully addressed, or if they have been addressed, the right answer has not been found:

- if Eva-f and -p preferentially occur at top lines, it could be that they are ornamental, i.e. nulls or a variant of other letters, or a combination of that. Can this be reconstructed? Note that there are some differences in the behaviour of f and p. Can this help to find out more?

- the unusual behaviour of line-initial words has not been quantified. There is some work from Emma, I believe, but this deserves a further look

- the unusual distribution of line-initial characters has not been quantified. This is likely to be one component in the previous question, but it won't explain everything.

- how to recombine glyphs in order to increase entropy? I did some very initial experiments with this that were quite promising, but the number of possible permutations is a big problem here.

- how can the difference between Currier A and B languages be better quantified? This has been looked at a lot, but no adequate answer exists so far. Note that I disagree with Nick, and don't consider the difference between A and B a major thing. For me it is relatively small, but the words "major" and "relatively small" are quite subjective of course

- should one look closer at languages not written in the Latin script (I think: yes). I think one should concentrate on those that would be available in late medieval Europe in MS form.

- the impact of spelling variants could be looked at, but such variants should only be expected for the vernaculars (a linguist should be able to say more about that).

The initial bullets look more at character issues, but any answer in those areas might help to provide new insights into the general topic of parsing.

[Aga Tentakulus]

I think spelling is an important factor to consider.
For example from a battle chronicle:
If powder was written as "pulver, bulver, pulfer, bulver". Also swords as swords "schwerter, swerter" the "ch" was simply omitted.
"pernere" I had to read more than the whole sentence until I realized that it is a person from the city of Bern.
Sometimes you can understand it well, we write ski, but speak schi.
And this from a chronicler, the same book, the same person.

As mentioned by Rene, the foreign text should be considered.