Automated Topic Analysis of the Voynich Manuscript

[quimqu]

Sorry to insist about my analysis, but I would like to share some thoughts about the previous plot with you, and if you wish to write your ideas, they would be very welcome.

I find this plot fascinating. Looking at it, from left to right, it seems that the MS starts with a very strong topic 0 (let's call it as Currier did A). Most of the first 1/3 of the herbal, it is almost pure A. But then gruadually, paragraphs that mix B topic words appear, and appear quite strongly. They are not as continuous as strong topic A paragraphs, but they appear suddenly after 1/3 of herbal paragraphs. Are the herbs starting to be described partially with topic B?

Then we get to the cosmological and astronomical part of the MS. There, we can see that there is a mixture of topic A and B. Note that at those folia, I "created" the paragraphs grouping the small sentences and labels, making a sort of bag of words. These words, even if they are mixed, seem to tend to topic A (look that the black line is above 0).

Then we get into the zodiac part of the MS. It starts above 0 and finished below 0. It seems that the mean topic changes from A to B gradually in that part of the MS. Well, it is maybe not so strange... maybe the stars apply to the plants and the zodiac apply to the humans. So there is a tend from topic A (plants) to topic B (humans) all through the astronomical and zodiac part.

Because when we get into the biological part, topic B is clearly the boss there. No place for topic A (the plants topic) but for topic B (the human? topic). 

But then, when we go further to the following cosmological part... look how different it is from the first cosmological part! It is clearly a topic B cosmological part, while the first one was more about topic A.

And what happens when we get into the Pharmaceutical part? Again topic A is the main topic. Strange? Well, not at all. If that part talks about herbs and recipes, maybe the herbal topic is the natural one.

But look again the change when we get into the marginal stars part of the MS. The main topic changes again to B. But look how mixed it is. There are plenty of paragraphs that have almost 50/50 A/B topic words. Maybe pages about medical recipes that mix topic A (herbal) with topic B (human)? Maybe medical treatments?

Well, I find it fascinating. As said at the beginning of this post, your thoughts are very welcome!

P.S: I know that this kind of machine learning algorisms might seem complicated or even strange to apply to the Voynich. I would suggest you to look at the results of a test Jorge Stolf proposed to me: see if that kind of models were able to distinguish Portuguese, phonetic Portuguese and Spanish. And they did it clearly. See You are not allowed to view links. Register or Login to view.to get an idea of how the models work and how to interpret the results on the Voynich (if we can iterpret them...).