[liblouis-liblouisxml] SV: Re: SV: Translation problems with some tables
- From: Bue Vester-Andersen <bue@xxxxxxxxxxxxxxxxxx>
- To: <liblouis-liblouisxml@xxxxxxxxxxxxx>
- Date: Fri, 24 Jul 2015 14:14:05 +0200
Hi Daniel,
You wrote:
Do you know how I can find out what tables don't yet support back
translation? Or is it just a case of testing each one individually.
You could ask on the list. Perhaps the author of a specific table knows if it
can be used for back-translation. Other than that, I think it is just "trial
and error", or hopefully in some cases "trial and success".
Can I assume that back translation should work for all computer Braille tables?
Probably not, but the chance is probably greater than with grade 2 tables.
However, computer tables usually only cover a very limited character set,
especially if it is a 6 dots table.
What is required to get the tables working with back translation? Is it just
a matter of update the ctb files? How much effort is generally involved in
implementing this and what's the best way to test it all > works correctly?
Hmm, complex question. In principal, it should be enough to modify the tables.
However, there may also be areas of the code that could do with some tender
love and care. For instance, I am not quite sure that word limits are treated
the same way in forward and backward translation.
In general, I would say that back-translation could be more difficult than
forward translation, because we have to mimick what a human is doing when
reading Braille. Also, there are a lot of pitfalls in the tables themselves.
For instance, tables that use context and pass2 a lot are more difficult,
because these opcodes currently have no effect when back-translating. Also,
some tables start out by including the file unicodedefs.cti. This means that
some or most dot combinations will be back-translated to there unicode
equivalent in the range \x2800-\x28ff. In stead, the file should be included
after the definition of all single cell characters.
Finally, some Braille codes are better suited for back-translation than others.
Some Braille codes are so ambiguous that you more or less have to know what you
are reading in advance. Also, you may lose some information during translation
that can't be recreated during back-translation, e.g. capital or accented
letters in some Braille codes.
There are lots of other things to be aware of when designing for
back-translation depending on the nature of the table in question, but these
are just a few general points.
Finally, do you know of any other projects that use liblouis for it's back
translation service?
NVDA does, but currently only computer Braille, not grade 2, perhaps because
many tables in liblouis don't support it.
I suppose BrailleBack for Android uses Liblouis for back-translation, but I
haven't tried it.
The Braille Sense note-takers from HIMS use liblouis for translation both ways,
but currently only in Danish. In other languages, I think they mostly use
Duxbury.
The GUI for Windows that I am developing will also support translation both
ways if the tables support it.
There may be other projects, but I can't think of any just now.
HTH Bue
Thanks,
Daniel
On Fri, Jul 24, 2015 at 6:22 AM, Bue Vester-Andersen <bue@xxxxxxxxxxxxxxxxxx>
wrote:
Hi Daniel,
I think that many of the tables in liblouis have not been designed with
back-translation in mind. Probably, the reason is that they were designed for
use with a library service or in a Braille embosser, so back-translation would
not be a big issue. However, I think it is something that we should take more
seriously as liblouis is getting more popular with both screen readers and
note-takers.
Currently, all the Danish tables have good back-translation. I am going to do
what I can over time to help improve back-translation for other tables as well.
However, I have quite a heap of projects on my hands at the moment. So I can't
make any specific promises.
Best regards Bue
Fra: liblouis-liblouisxml-bounce@xxxxxxxxxxxxx
[
mailto:liblouis-liblouisxml-bounce@xxxxxxxxxxxxx] På vegne af Daniel Dalton
Sendt: 23. juli 2015 15:59
Til: liblouis-liblouisxml@xxxxxxxxxxxxx
Emne: [liblouis-liblouisxml] Translation problems with some tables
Hello,
I'm writing an app that relies on loui for back translation. It works fine with
the English tables. However, having trouble with Polish and Spanish. Firstly I
do not use either of these codes, so they are issues reported by users, but I
can reproduce. They are having trouble typing punctuation like . , ? !
For example in Polish if I do lou_translate from my console on Linux forward
translation a "." should be in Braille as ' or dot 3. If I then forward
translate this ' it just returns '
In Polish when I back translate some letters like a b c through j it would
return the numbers 1 2 ... 9 and 0. I fixed this by making the change of
removing the include digits6Dots.uti and replacing it with litdigits6Dots.uti.
Not even sure if this is correct.
Finally, when I programatically back translate with the java wrapper I some
times get back the dot number where something has no corresponding translation.
For instance if I pass dot 7 by its own and it has no corresponding symbol in
that braille table it'll return "7". Will I just need to write my own logic to
not show this to my user?
Thanks in advance for any help with any of this.
Daniel
Other related posts:
