Re: [lojban] Computer Friendly Handwriting

a single letter, on unlined paper is the probably the one possible scenario where it does have a problem, but that is easily solved.

you could use the ratio between line segment lengths (with a clear vertical range ratio is moot, horizontal lines can be longer or shorter than the vertical range with no effect of data... if you wanted to draw one character and scan it all by itself, just ensure the ratio can be used to determine this)

eg
vertical line is either the same length or shorter than the horizontal
or is longer/ double length

The greater problem for "single character scanning" in my mind would be rotation, what if it is upside down, but this is also a problem for standard latin alphabet (p vs d, q vs b)

im my mind there are three possible cases

1)human written on lined paper - No Problems

2)computer generated no line - No Problem (ratio would be perfect, and computer can generate perfect alignment without lines)

3)human written no lines - difficult and minor Problems

The minor problem is identifying the range, especially if your don't write in a straight line or write only a few characters that all happen to the same height

It would be easy enough to just draw a vertical line before the text t establish the range (eg. "| cig"). Or any other method where a marking is meant to establish a unit of length or height to be referred to

as per "sloppy handwriting". this problem exists with ALL computer reading of human written text, NO system can fix this

even human have trouble reading each others sloppy handwriting

The question should be "can it read sloppy all handwriting"?

but instead "what is the margin of error? how easily are errors made while writting? and how does this compare to other formats?"

without proper study i cant say.. but I think it is safe to guess Cscript will outperform latin script on lined paper

basically it is designed for lined paper

while it CAN be written by hand without lines, it is not optimized for that use

don't have lined paper???

just grab something straight and draw your own lines ;)

I plan to actually use the lines in the first versions of the scanning software, first i will scan for long line, establish rows, and scan them, much more accurate and faster than trying to auto-detect

On Tue, Jul 22, 2014 at 10:29 PM, 'Seth Turner' via lojban <lojban@googlegroups.com> wrote:

i was thinking without lined paper, written by a person. computer should have a very easy time with this stuff though

On Jul 22, 2014, at 5:22 AM, Matthew DeBlock <vastencn@gmail.com> wrote:

well using lined paper it should be pretty easy (middle vs. touching top or bottom)

also, assume top =100 bottom=0
the difference in hight can be more than half

eg.
0,1,1 would be 0-50-50
0,2,2 would be 0-100-100

but 011 could also be written 0-10-10
this would make the difference even greater adn easier to see (it woul dnot be confused with 000 because 000 would not have any corners)

if you mean for the computer reading..
then you are perhaps confusing this with glyph based OCR
the computer would never looks ta the "shape as a whole".. that is needed for glyph recognition(like latin alphabet) but that whole level of the software is not needed here

basically dont think of them like you would normal glyphs, you would never "compare shapes", just read "absolute hight values"

On Tue, Jul 22, 2014 at 1:07 PM, 'Seth Turner' via lojban <lojban@googlegroups.com> wrote:

I'm not sure how you are supposed to tell the difference between letters of the same shape at different heights, like a and n, d and q. I'm otherwise quite a fan of all of your work

On Jul 21, 2014, at 9:16 PM, Matthew DeBlock <vastencn@gmail.com> wrote:

Of course I will.

Ill try to remember to send you a copy of the code when I get around to it :)

On Tue, Jul 22, 2014 at 3:41 AM, .arpis. <rpglover64+jbobau@gmail.com> wrote:

(Suppressing snobbery about language choice)

That sounds really cool! I hope you'll post on your facebook page when it's done.

On Mon, Jul 21, 2014 at 8:54 AM, Matthew DeBlock <vastencn@gmail.com> wrote:

as per "underestimating" OCR requirements.
as I pointed out this would be more like QR codes, it only has to "trace" the line and it has all the data. OCR needs an extra step where it compares likeness to known glyphs. a whole leve lis gone.

I plan to build some readng software as soon as I have time. I am planning on first doing it in PHP so it is easy for others to test and modify, but to do a reall "speed test" I will need to do it in C

As per Palm OS graffiti
That only works if it is draw into the system in real time. not if it is scanned in after the fact. it relies on knowing the stroke direction, something that is lost in scan of written text.

On Mon, Jul 21, 2014 at 1:16 PM, Robin Lee Powell <rlpowell@digitalkingdom.org> wrote:

Rather the opposite; http://en.wikipedia.org/wiki/Graffiti_(Palm_OS)
worked perfectly on CPUs that now probably are used to drive
*watches*; a modern smartphone is many many times more powerful.

My question is why the hell you would invent something as painfully
terrible on the human as Cscript when Graffiti is know to be
entirely workable in real time on what is in the modern day a
trivial processor.

On Sun, Jul 20, 2014 at 03:17:17PM -0400, .arpis. wrote:
> Not to be unreasonably pessimistic, but I will believe that the design goal
> of computer legibility to be satisfied when I see a program that
> demonstrates it. Both Cscript and Dscript are cool, and I'll definitely
> play around with them on my next con-script kick, but I worry that you're
> underestimating the difficulties of OCR on real text written at a
> reasonable speed.
>
>
> On Sun, Jul 20, 2014 at 8:25 AM, vasten <vastencn@gmail.com> wrote:
>
> > A logban user contacted me and says he is using this for lojban, figured
> > you guys might like too :)
> > [ http://www.facebook.com/Dscripting - see comments in the top post on
> > this page for details ]
> >
> > He says tendinitis in the right hand forces him to use his left hand and
> > he find Cscript easier.
> > Seems he has mapped it to lojabn quite easily.
> >
> > CSCRIPT - Computer / Human Bi-Friendly Writing System
> >
> > http://dscript.ca/cscript.pdf
> >
> > Cscript is designed to be both easy to read and write by humans as well as
> > be digitally and programmatically read and written by computers.
> >
> > There is of course a trade off between the two. Cscript could be thought
> > of as "lying somewhere between QR codes and standard hand writing"
> >
> > For Humans it easy to produce with standard lined paper and allows some
> > intuitive "cursive" elements.
> >
> > For Computers it removes entirely the need for glyph recognition and
> > shape/vector analysis. It could read straight off the raster level as a
> > string of absolute values. It eliminates the need for an entire level of
> > OCR shape comparison, and drastically reduces ambiguity, requires FAR less
> > processing power and increased accuracy.
> >
> > It is not meant to be "perfect" for either, but instead offer a more
> > balanced alternative.
> >
> > **technically it would not be considered "OCR" as it does not actually
> > require the "character recognition" level of the software at all.
> >
> > ***keep in mind the "value range zones" can even be flexible based on
> > context (ie. assume top=1, bottom=0.. [0->0.33]=0, [0.34->0.66]=1,
> > [0.67->1]=2... the values [0 , 0.1, 0.1] = "0-1-1" because there would be
> > no corner point in 000
> >
> > Another new project of mine that might be fun too....
> >
> > CHEMICAL CALLIGRAPHY
> >
> > http://dscript.org/chem.pdf
> >
> > This is a mnemonic device and art form.
> >
> > It is designed to allow simpler representation of bio-chem molecules with
> > "less noise".
> >
> > It drops some information, which is assumed is "obvious" to someone with
> > basic chemistry knowledge. The missing information can usually be "filled
> > in" with basic chemistry understanding.
> >
> > It allows various forms of any one molecule (the larger the molecule the
> > more possible forms), which adds greatly to the users ability to make
> > artistic and aesthetic choices without altering the molecular and
> > structural information.
> >
> >
> >
> > Still have not rebuilt my lab, current apt too small :( So still just
> > hacking with "pen and paper" ;)
> >
> > Hopefully soon will be able to get back to my mad science :)
> >
> > --
> > You received this message because you are subscribed to the Google Groups
> > "lojban" group.
> > To unsubscribe from this group and stop receiving emails from it, send an
> > email to lojban+unsubscribe@googlegroups.com.
> > To post to this group, send email to lojban@googlegroups.com.
> > Visit this group at http://groups.google.com/group/lojban.
> > For more options, visit https://groups.google.com/d/optout.
> >
>
>
>
> --
> mu'o mi'e .arpis.
>
> --
> You received this message because you are subscribed to the Google Groups "lojban" group.
> To unsubscribe from this group and stop receiving emails from it, send an email to lojban+unsubscribe@googlegroups.com.
> To post to this group, send email to lojban@googlegroups.com.
> Visit this group at http://groups.google.com/group/lojban.
> For more options, visit https://groups.google.com/d/optout.

--

http://intelligence.org/ : Our last, best hope for a fantastic future.
.i ko na cpedu lo nu stidi vau loi jbopre .i dafsku lu na go'i li'u .e
lu go'i li'u .i ji'a go'i lu na'e go'i li'u .e lu go'i na'i li'u .e
lu no'e go'i li'u .e lu to'e go'i li'u .e lu lo mamta be do cu sofybakni li'u

--
You received this message because you are subscribed to a topic in the Google Groups "lojban" group.
To unsubscribe from this topic, visit https://groups.google.com/d/topic/lojban/tP3LTTlV3x0/unsubscribe.
To unsubscribe from this group and all its topics, send an email to lojban+unsubscribe@googlegroups.com.

To post to this group, send email to lojban@googlegroups.com.
Visit this group at http://groups.google.com/group/lojban.
For more options, visit https://groups.google.com/d/optout.

--
You received this message because you are subscribed to the Google Groups "lojban" group.
To unsubscribe from this group and stop receiving emails from it, send an email to lojban+unsubscribe@googlegroups.com.
To post to this group, send email to lojban@googlegroups.com.
Visit this group at http://groups.google.com/group/lojban.
For more options, visit https://groups.google.com/d/optout.

--
mu'o mi'e .arpis.

--
You received this message because you are subscribed to a topic in the Google Groups "lojban" group.
To unsubscribe from this topic, visit https://groups.google.com/d/topic/lojban/tP3LTTlV3x0/unsubscribe.
To unsubscribe from this group and all its topics, send an email to lojban+unsubscribe@googlegroups.com.
To post to this group, send email to lojban@googlegroups.com.
Visit this group at http://groups.google.com/group/lojban.
For more options, visit https://groups.google.com/d/optout.

--
You received this message because you are subscribed to the Google Groups "lojban" group.
To unsubscribe from this group and stop receiving emails from it, send an email to lojban+unsubscribe@googlegroups.com.
To post to this group, send email to lojban@googlegroups.com.
Visit this group at http://groups.google.com/group/lojban.
For more options, visit https://groups.google.com/d/optout.

--
You received this message because you are subscribed to a topic in the Google Groups "lojban" group.
To unsubscribe from this topic, visit https://groups.google.com/d/topic/lojban/tP3LTTlV3x0/unsubscribe.
To unsubscribe from this group and all its topics, send an email to lojban+unsubscribe@googlegroups.com.
To post to this group, send email to lojban@googlegroups.com.
Visit this group at http://groups.google.com/group/lojban.
For more options, visit https://groups.google.com/d/optout.

--
You received this message because you are subscribed to the Google Groups "lojban" group.
To unsubscribe from this group and stop receiving emails from it, send an email to lojban+unsubscribe@googlegroups.com.
To post to this group, send email to lojban@googlegroups.com.
Visit this group at http://groups.google.com/group/lojban.
For more options, visit https://groups.google.com/d/optout.

--
You received this message because you are subscribed to a topic in the Google Groups "lojban" group.
To unsubscribe from this topic, visit https://groups.google.com/d/topic/lojban/tP3LTTlV3x0/unsubscribe.
To unsubscribe from this group and all its topics, send an email to lojban+unsubscribe@googlegroups.com.
To post to this group, send email to lojban@googlegroups.com.
Visit this group at http://groups.google.com/group/lojban.
For more options, visit https://groups.google.com/d/optout.