A few remarks and notes:
- The resolution, close to 50% of the Voynich-Final-Sequential_ID (2014) series, is good enough for web browsing, much better than voynichese.com.
- It must have been a lot of work to make the new bounding-boxes (and sometimes rotated boxes and even polygons, that's new and an improvement over voynichese.com). Voynichese.com used the 2004 scan and the coordinates cannot be converted because of perspective: pages aren't flat.
- The transliteration system is interesting and logical. I haven't understood everything but it seems that C and H work like { and } in EVA and " works like ' in EVA so there
should be a way to convert between the two formats, I hope.
c stands for Ce
h stands for eH
s stands for c"
(18-02-2023, 10:31 AM)ReneZ Wrote: You are not allowed to view links. Register or Login to view.This is very impressive work indeed.
It looks like this is mostly a new transliteration.
It is different from the Takeshi file in very many places.
Did you make it yourself?
I took transliteration from You are not allowed to view links.
Register or
Login to view. as a starting point (which is probably one of Takahashi), then made some adjustments (mostly correcting mistakes and redefining complex symbols and ligatures). I couldn't identify the creator of voynichese.com, so it was done without explicit permission.
(18-02-2023, 01:45 PM)nablator Wrote: You are not allowed to view links. Register or Login to view.A few remarks and notes:
- The resolution, close to 50% of the Voynich-Final-Sequential_ID (2014) series, is good enough for web browsing, much better than voynichese.com.
- It must have been a lot of work to make the new bounding-boxes (and sometimes rotated boxes and even polygons, that's new and an improvement over voynichese.com). Voynichese.com used the 2004 scan and the coordinates cannot be converted because of perspective: pages aren't flat.
- The transliteration system is interesting and logical. I haven't understood everything but it seems that C and H work like { and } in EVA and " works like ' in EVA so there should be a way to convert between the two formats, I hope.
c stands for Ce
h stands for eH
s stands for c"
Unfortunately full size images are currently too heavy for online app, so halved images is reasonable compromise for now.
Coordinates were also taken from You are not allowed to view links.
Register or
Login to view., then mapped to bigger images (this is tricky process, yes: you should find subimage in scaled and rotated image).
Transliteration in the end is based on simple rules: small letters are standalone, upper case letters are to be combined with small (like diacritics), digits are versions, ' is upper case version of ".
s
stands for c", but c
can be considered as Ce. The difference is that when you search for c or c", you naturally find also s, but when looking for e it may be confusing to yield also c or h.
It's beautiful!
Thank you.
(18-02-2023, 04:14 PM)farmerjohn Wrote: You are not allowed to view links. Register or Login to view.I took transliteration from You are not allowed to view links. Register or Login to view. as a starting point (which is probably one of Takahashi), then made some adjustments (mostly correcting mistakes and redefining complex symbols and ligatures). I couldn't identify the creator of voynichese.com, so it was done without explicit permission.
I converted the transliteration used on voynichese.com to the IVTFF format, and it is indicated by the abbreviation "VT" in You are not allowed to view links.
Register or
Login to view. . It turns out to be identical to the IT transliteration, with all cases of '?' (unknown reading) removed.
The IT transliteration is Takeshi Takahaki's, as included in the Interlinear file.
Having quickly gone through this document: You are not allowed to view links.
Register or
Login to view.
I have the impression that there is no "quick" way to convert the alphabet used here to Eva or back.
The following 'almost' works:
c -> {c
C -> {
h -> h}
H -> }
One issue is that probably not all variant characters exist in extended Eva. So far I have only
seen q1 as such an exception, but there may be more.
The bigger problem is that some character translations depend on context. The clearest case is the single
e
which converts to Eva-e when it is outside a ligature or to Eva-h when it is inside.
Both problems can be solved in a rigorous way by going to the STA alphabet, using long conversion tables.
However, the result may not be representable completely either in extended Eva or v101.
(I am having a similar issue with a new transliteration which is an automated merger of the ZL and GC files.)
Some more updates.
Web version of the browser/search with full size folio images has been added. There are certain concerns about its usability, since normal folio image weighs almost 10M, but Rosettes image is more than 45M.
sh is now transliterated as csh, so s now stands for "the plume" only. This releases symbols ' and " and solves several issues, the most awkward one being the bug with my font on iPhone (caused by the implementation of s symbol).
(24-02-2023, 12:17 PM)farmerjohn Wrote: You are not allowed to view links. Register or Login to view.Some more updates.
Web version of the browser/search with full size folio images has been added. There are certain concerns about its usability, since normal folio image weighs almost 10M, but Rosettes image is more than 45M.
sh is now transliterated as csh, so s now stands for "the plume" only. This releases symbols ' and " and solves several issues, the most awkward one being the bug with my font on iPhone (caused by the implementation of s symbol).
Hi Konstantin,
On Firefox only, the browser does not start any more, with the error "Uncaught ReferenceError: urlParams is not defined" in the console.
This line was removed since last week:
const urlParams = new URLSearchParams(window.location.search);
(24-02-2023, 03:39 PM)nablator Wrote: You are not allowed to view links. Register or Login to view.Hi Konstantin,
On Firefox only, the browser does not start any more, with the error "Uncaught ReferenceError: urlParams is not defined" in the console.
This line was removed since last week:
const urlParams = new URLSearchParams(window.location.search);
Hi
nablator,
Please try clearing cache or opening the page in private mode.
It's working now that I've cleared the cache. Thanks!
I love the browser. Bookmarked.
---
There is a small mistake on f23v: qoteotor should be qotcoHtor
(18-02-2023, 01:45 PM)nablator Wrote: You are not allowed to view links. Register or Login to view.Voynichese.com used the 2004 scan and the coordinates cannot be converted because of perspective: pages aren't flat.
I have been working in this area off and on, and it came as a bad surprise that a straightforward coordinates conversion between the old and new scans was not going to work. I can now also confirm this, after having given it a try. However, I did find a way to do it, as follows.
First, the new Beinecke scan is also reduced in size to have a height of 1500 pixels.
Then, for a given page, I take the centre of each of the word boxes, and take a small area around it from the
old scan. This is converted to lightness (inverted, i.e. there will be light text on a dark background).This would look like this:
[
attachment=7247]
Then, I take a larger area of the
converted new scan, and I produce the same. That would look like this:
[
attachment=7248]
Then I move the smaller plot over the larger plot and find the point where the correlation of the lightnesses is highest. The two can be shown together, as below. The lightness of the larger area is in the blue channel and that of the smaller area is in the red channel.
[
attachment=7249]
This process provides the offset between the two images. Then, the result can simply be scaled up for the full-size Beinecke scan.
I tried this on two pages so far, and am happy with the result.
Of course, there may be extreme cases where this could still fail.