(31-10-2023, 04:02 PM)Koen G Wrote: You are not allowed to view links. Register or Login to view.It's an interesting and well done talk. However, I am currently a bit over halfway through and I see a problem, which is that Currier languages are ignored. So he discovers a statistical shift around f33, which is where we also see a shift towards Herbal B. To me this feels kind of like discovering where a new 'language' starts, much more than a new topic or a new author. (Even though we do know Currier languages also correspond to different hands). The analysis should really have been done separated by Currier language, in my opinion.
I think that Mark made a similar point during the questions at the end, particularly with respect to changing hands. I'm definitely interested in following this line of thought (when time allows). I think it would be of interest if the statistically identified shift (or shifts, if I run a multiple changepoint analysis) were correlated with otherwise identified shifts in writing style.
As I mentioned in the talk, and in the earlier post, I'm definitely aware of the extent of the literature around this and the limits of my knowledge of that literature. So I'm always happy to be linked to things that put the statistical findings in context!
I'm glad you enjoyed the talk.
I found particularly interesting the presented analysis by means of stylometry / stylo-statistics) ( a method which was completely unknown to me so far. ). The frequency of word lengths is considered here as a stylistic indicator for the writer. There is a noticeable change very early in the VMS in folio 33. It is hard to say whether this study could be "language dependent" or not.
I guess what I mean is that there is a bit of a mismatch between the advanced methods used and the pretty obvious differences between Currier languages (like the f33 shift). You explain in the video how certain statistical methods can be used to detect differences we may not notice by eye, which I found very interesting. For example, we can see if a new author finished a book, and where the input of the new author starts. So while language, genre and subject stay the same, the method is still capable of picking out the new author's quirks.
But Currier "languages" have been known for decades and are clear enough to be noticed by someone without using a computer. So noting a shift in f33 is kind of like saying you used topic modeling to determine that French isn't Italian. In my opinion it would be much more interesting to determine where the shifts are within the same Currier languages, whether the different scribal hands identified by Lisa Fagin Davis can be seen in the statistics etc.