Elian Script - enhancements

Please note that this posting is in my "Half-baked ideas" category and is really a collection of notes on something not yet finished. I'll add images and code as the idea progresses.

I've always been interested in shorthand or some other way of improving the speed and readability of my handwriting. It's probably some sort of psychological tick that comes from receiving a "Needs Improvement" on my report card for pretty much every single year that such things were tracked during my education.

There have been several things that have bugged me about most of the systems that I've looked at:

• Are transcriptions of sounds rather than words. Because of the way I read and type, I really don't deal with my written language in sonic units. So, when they don't actually spell out the word, I have to think more about them and thinking more about a system just leads to its being dropped from my use.
• Similarly are those that rely heavily on abbreviations and omitting vowels, etc. In many of these systems, if you don't get the notes transcribed into a full text using regular Latin script, you are pretty much screwed. I don't want notes that can't even be read by someone who understands the system.
• Are overly reliant on strict form. Slight deviations in forming the glyphs results in being another character. Given my clear inability to form the letters I've been writing for 25 years properly, this is a non-starter.

So, when I discovered ElianScript, I was excited. It not only looked cool, but was a letter for letter transcription of the Latin alphabet and the forms are easily distinguished, even if formed sloppily.

It's based on a 9 unit grid with each glyph being the interior facing lines for that glyphs position in the grid. 3 modes give 27 glyphs. Each mode extends the previous mode's glyphs. In the 1st mode, all lines are equal, giving basic portions of even squares. In the 2nd mode, one of the lines is extended. It doesn't matter which one (this gives some real flexibility which will come into play for my modifications). In the 3rd mode, a dot is added to the end of one of the lines in addition to one of the lines being extended.

Ms. Elian built her script by alphabetically distributing the glyphs through the 3 modes, a-z, starting with "a" at the top left of mode 1. Her primary purpose in the system was to deliver a writing system for her calligraphy work (which is quite beautiful). For that purpose, it works wonderfully. The forms have enough variety to let her explore language and word forms extensively.

However, for a more general writing purpose, I think it can be improved. Since she's primarily using it for art, the practicality isn't really needed for her purposes. However, since I would like to use it for notetaking and more general handwriting, I'd like to suggest a few improvements.

My improvements do not detract from the potential in art. The same glyph grid exists with my modifications. Each glyph can still be used in it's various forms.

So, if I'm not changing that stuff, what am I changing?

• A preferred form of each glyph
• An English optimized distribution of the letters into the grid.
• An elimination of the 2nd line glyph in grid 3 (3-2 as the skipped glyph instead of 3-3).

Overall, these changes are to fulfill a couple of goals (and give me even more lists in this posting):

• English is written from left-to-right. 20+ years of experience in this mode can be borrowed against to make it efficient. So, glyphs ending in a position where you can easily start the next is preferred.
• The number of pen strokes should be minimized. The fewer times you have to change direction and move the better.
• The best setup is one where the strokes can be fluid, yet easily recognized when "messed up" a little bit.
• If vowels are all grouped together and use the simplest, smallest glyphs, you'll end up with a nice rhythm to the writing that fits with non-Latin systems (i.e. vowels as "minor" letters) without actually getting rid of them or demoting them to accents only.
• Since both existing paper products aimed at writing Western languages and mindsets are organized around a core line height with ascenders and descenders, preferring vertical forms of glyphs over the wide form makes sense and will likely eliminate wasted space. The rising and descending tend not to interfere with each other in regular Latin-based English, I suspect this will be the case here and result in an efficient setup.
• The final stroke should be the long one. Accelerating off into the page is much easier than deccelerating to a specific point where you have to turn. When it doesn't make sense for it to be vertical, an angle under the rest of the line, toward the right should make sense.

OK, so I've got noble intentions. Now, how do I go about optimizing this thing?

Build a harness for seeing what the results will look like. This is done by taking the rigid form glyphs from her site, as is, making individual characters out of them, generating a grid, assigning glyphs to letters and running test strings through. A further set of graphics needs to be made with the optimized glyphs to make this tool better.

Each grid can be saved and compared through iterations. Unfortunately, I can't see a way to run this through a genetic algorithm easily (like the typing layout stuff out there) as the parameters aren't easily read by a computer.

The grids will be generated using the frequency tables from LetterFrequency.org. In addition to basic letter frequency, common words, trigraphs and doubled letters will be included in test runs to ensure that things like "qu" are attractive and easy to do.

The resulting tweaks should give an optimized glyph mapping with preferred glyph forms for quick notetaking. While my set will be optimized for English, this same methodology should work to generate optimized mappings for other languages. Similarly, a 4th mode or modification to characters could be added for accents. My first thought on this is a "cross" like a Latin "t". Since the script already uses dots, the cross as a modification makes sense.

Once I get a theoretical mapping and set of preferred glyph forms, I may try to find a calligrapher to do a nice set of alphabets for reference and examples of the fluid form.

A font is likely going to be difficult given the level of efficiency obtained in the existing samples for overlaps and nesting of characters. It's very much like Korean in this regard, where a word is nested together and read like a clock face as a single glyph.

Then, I can see how quickly I can learn it for my own purposes.

[edit: 7-24] - I've completed my first pass at remapping. I used the letter frequency count for general fiction. Vowels first, in order except for "o". It didn't make sense to have the complete box glyph be the "u" when they were so close together. I may have to do the same thing with "t" and "u". So far, it looks OK for the "lorem ipsum", etc.

Given the deficiencies in the general fiction list (no "b" for some reason), I grabbed the "general English distribution" as my working list of glyphs.

Here's an example of my current mapping, using the default fluid glyphs (not rotated or adjusted for speed).