Idea to find the word Root, Leaf, Stem, Flower & Seed.

[nablator]

It found 33 total occurrences. 33 occurrences ÷ 111,325 tokens × 1,000,000 = 296.43.

Correct, thanks. Edition here: You are not allowed to view links. Register or Login to view.

The values for Macer floridus are also correct.

root should be fairly obvious to spot

So how do you spot the word that means "root" if you don't know that it is radix/radic* in Latin? Many words are more likely to occur in herbals than astrology books.

[R. Sale]

Ready, set, have a go at the VMs. Do the results make any sense? It's not the inability to analyze, it's the difficulty in the interpretation.

If you're looking for multiple repetitions in a short VMs text sequence, try the outer ring of White Aries, compared to any of the other circular text bands in the VMs zodiac sequence.

[npcompl33t]

It found 33 total occurrences. 33 occurrences ÷ 111,325 tokens × 1,000,000 = 296.43.

Correct, thanks. Edition here: You are not allowed to view links. Register or Login to view.

The values for Macer floridus are also correct.

root should be fairly obvious to spot

So how do you spot the word that means "root" if you don't know that it is radix/radic* in Latin? Many words are more likely to occur in herbals than astrology books.

You are correct, but I need to clarify a few points.

First, root was considerably more common than other botanical terms in most herbals, by a fairly significant margin.

Second,  I did not limit the search to just root. There are also other features in the image that can be used to distinguish other terms. For example, is a flower depicted, if so, flower is expected to be more common.

Third, an exact identification is not necessary to be useful. If a list of possible candidates is produced, you can start testing identifications and looking to see if the words themselves seem to offer clues. Essentially a list of possible cribs.

Fourth, and this is the largest issue: Doing this type of analysis, there is generally not many actual candidates for these words across both Herbal A and Herbal B. Like I mentioned, A lacks a valid non herbal comparison, but If both texts genuinely are herbals, there should be a shared botanical set of words common in both , that really don’t seem to exist. 

Final the paper I shared on blind machine translation essentially does this for all possible words. The idea is to learn expected contextual appearances and frequencies of words, assigning one at a time to try to match that of the training data. In this case, training on other herbals.

So yes, simply trying to assign the word root is not likely to be successful, but collectively, across all botanical terms, with the accompanying drawings, and historical references, finding enough information to generate a shortlist of possible candidates, is a much different problem.