Strong evidence of a structured four-phase system in the Voynich Manuscript

[Urtx13]

It seems the manuscript has internal thematic organization beyond random noise.

It was suggested before. Personally I never got deep into it but some research shows that the pages seem to make some information clusters. Still it doesn't mean that the text is meaningful. The scribe could just repeat together some nonsense words through several pages and then drop them and switch to another ones.

The problem is that we don't really see any grammar in the text - sentences, structures, words working as sentence starts, verbs or conjunctions...

Anyway I'll wait for your article because at this moment we just don't have sufficient info for further discussion.

Yes! 

Some previous studies indeed suggested that there might be clusters in the Voynich manuscript, based chiefly on surface clustering or visual observations.

However, my current approach is different:
-It starts from clean EVA tokenization (and later cross-checking with Takahashi),
-Applies both TF-IDF clustering and LDA topic modeling,
-Tracks thematic transitions across the manuscript,
-Compares the Voynich against multiple random control texts to validate the findings statistically.

Thus, it’s not just clustering pages based on similarity — it’s a complete pipeline to test structural organization rigorously.

Of course, proving true “meaning” or grammar would require much more, but the first step is to demonstrate that there is a non-random structure.

I understand it may not be very clear...

Many people are naturally drawn to the text itself, but from a linguistic perspective, detecting internal structural patterns is crucial.  It gives us hints about whether the text was intentionally organized, and about how we might eventually approach decoding the language.

It is doubtful that the Voynich manuscript will ever be “translated” from scratch without finding internal structure; there will be no sudden Rosetta Stone. That is a romantic idea.

I’ll share the detailed results soon.