How could we approach the Pharma and Recipes?

[Anton]

In fact what we need is full transcription of a sufficient number of supposedly "relevant" texts, preferably in different languages (to see if there are patterns inherent to all languages).

You are not allowed to view links. Register or Login to view., Biblioteca Universitaria di Bologna MS 2861, pointed out by Nick looks like a good candidate (mostly Italian, some Latin, XV Century).

Yes and no. I mean not the source itself (which is a good candidate), but the transcription. This transcription is made to read and understand the text, not to analyze it from the statistical point of view. I think that, from methodological perspective, we should analyze contemporary writings from the same angle that we do analyze the VMS - that is, we should take words (tokens) at their face value. For example, if a word reads "uñ" (short for "und"), we should take it as "uñ", not expanding that to "und". Firstly, not doing so will distort character statistics. Secondly, the same word can be abbreviated in several ways, and those will be different tokens. Expanded form is one and the same token. That distorts the word statistics.

I approach the transcription as if I would do with the VMS. And I face exactly the same problems (my weak knowledge of German adds to that): is it one glyph or another? Is there a space there or no space? Should I attribute the macron to one letter or to two letters?