Confirmation that spaces are really spaces

[Koen G]

Do spaces ever convey any information that could be required for interpreting the text, or are they wholly predictable?

My question would be: does the predictable writing system cause predictable spaces, or do spaces make the writing system more predictable? In the first case, spaces are spaces. In the second case, they may be something else.

it’s likely no coincidence that the glyph pairs with the highest incidences of unexpected behavior also tend to have relatively high proportions of ambiguous word breaks represented in the Zandbergen transcription, i.e., what I call “comma breaks,”

Might this be caused by transcribers feeling less certain about this situation than about, for example, an uncertain space after [n]?

[quimqu]

If the spaces were decorative or had no function, the model should not be able to recover them with this level of accuracy.

On the contrary, the more predictable they are, the less information they can carry. 100% predictable = 100% useless.

It wasn't easy at all to find a model that detects morphemes that match spacing. As I explained, I started training models with natural languages, so I could check their accuracy. I must say that at the beginning it was almost impossible to get good results. The models tend to reduce the morphemes to a single char and then it makes nonsense. Think that the model starts with the text in a single line... I really think it is not that evident.

[rikforto]

Do spaces ever convey any information that could be required for interpreting the text, or are they wholly predictable?

My question would be: does the predictable writing system cause predictable spaces, or do spaces make the writing system more predictable? In the first case, spaces are spaces. In the second case, they may be something else.

Can you unpack this a bit? I am not intuiting the practical distinction here.

[Koen G]

Let's simplify and say that some glyphs or glyph clusters only appear at word end. Think n, edy... Because for some reason these dominate the ends of words, they also make the occurrence of spaces predictable, because a space follows word end. In this case you could say that spaces are perfectly normal: they are only predictable because the "words" themselves don't provide many different contexts for spaces to appear in.

In the second scenario, blank space is inserted in a predictable way. It is the use of blank space that makes "words" appear even more predictable than they were.

The result may be exactly the same, but the underlying system and the way we may think about spaces is different.

[quimqu]

Just to make it clear: the model finds morphemes by analysing a text without spaces (so all the words together). Then, to check the spacing, I compare the "real" delimited words with the morphemes that the model has found for them. If the morphemes fit exactly into the original word, the spacing at both sides is considered correct. The model does not find words.

[Philipp Harland]

Let's simplify and say that some glyphs or glyph clusters only appear at word end. Think n, edy... Because for some reason these dominate the ends of words, they also make the occurrence of spaces predictable, because a space follows word end. In this case you could say that spaces are perfectly normal: they are only predictable because the "words" themselves don't provide many different contexts for spaces to appear in.

In the second scenario, blank space is inserted in a predictable way. It is the use of blank space that makes "words" appear even more predictable than they were.

The result may be exactly the same, but the underlying system and the way we may think about spaces is different.

Predictable as in some entropy function is minimized to a certain extent? Am I getting this right?

[rikforto]

The result may be exactly the same, but the underlying system and the way we may think about spaces is different.

Ahhh, okay, I was too locked into thinking about the surface presentation, thank you

[Mark Knowles]

SpacesarenotessentialtomakesomethingreadablealthoughIthinktheyhelp

[ReneZ]

The spaces in the Voynich MS text very much blend in with the word pattern, or one could also say: they are part of the word pattern.

It is impossible (I think) to decide between cause and effect, and perhaps it is all part of a single system.

What I mean is: spaces tend to follow a certain group of characters, all of which are i- or c- shaped with a swirl. Now is the space there because of this character, or does the i or c have a swirl because it is end-of-word?

What really interests me is if the character groups between spaces represent words, or less than words.
The analysis presented here does not seem to say anything about that.

[quimqu]

The spaces in the Voynich give a superficial pattern. But the model takes as input the text without spaces or line breaks and identifies morphemes. These appear with internal consistency, showing stable distributions and transitions. 

These morphemes group together and separate in ways that do not always match the spaces (we could have a morpheme created from the end of one word and the beginning of the next). With the model, a single “word” can be divided into different morphemes depending on its surrounding context. This seems to indicate that the real structure of the text is richer than the simple sequence of “characters up to a space.” 

The model detects units that repeat reliably and behave like self-contained blocks. I believe the model can indeed provide a different perspective on word creation based on the morphemes it finds.