But couldn't you do that pre-processing in a correct pass, if that would
become the very first pass (and dots-to-dots)?
2017-01-11 11:01 GMT+01:00 Bue Vester-Andersen <bue@xxxxxxxxxxxxxxxxxx>:
Hi,
I, for one, need to use pass2 before the actual backtranslation. One
example is to correct the way capsword is used and what characters cancels
it.
So, I would have real problems, if pass2-3-4 were put after the
back-translation in stead of before.
Apart from that, I agree that the same multi-pass rules would probably
almost never work for both forward and backward translation. This is why I
am currently splitting up my 6-dots tables in tables for forward and
backward translation, and I will probably have to do the same with 8-dots,
unless nofor and noback get to work with the multi-pass opcodes.
Bue
*Fra:* liblouis-liblouisxml-bounce@xxxxxxxxxxxxx [mailto:
liblouis-liblouisxml-bounce@xxxxxxxxxxxxx] *På vegne af *Bert Frees
*Sendt:* 11. januar 2017 10:49
*Til:* liblouis-liblouisxml@xxxxxxxxxxxxx
*Emne:* [liblouis-liblouisxml] Re: SV: Corrections pass during back
translation.
I wonder if it would make more sense to apply backwards pass2, 3 and 4 in
that order and after the regular rules instead of before. This way they
would become "cc" (character-to-character) rules. They could then deal with
cases where back translation gets it wrong.
This of course only makes sense if the backwards passes 2-4 are seen as
completely separate passes specific to the backward translation, and not as
exact counterparts of the forward passes. So the question is: do forward
passes 2-4 need exact counterparts? In other words, do we really want to
perform the backward translation through exactly the same path as the
forward translation but in reverse?
I don't think that context rules are applied in backward translation. That
wouldn't make any sense would it? (Looking at the code confirms this.)
2017-01-11 10:35 GMT+01:00 Bue Vester-Andersen <bue@xxxxxxxxxxxxxxxxxx>:
Hi,
I agree with Dave. One example where the correct opcode is useful during
back-translation is words with accented letters. In many languages, foreign
letters with unusual accents are marked with a special accent mark.
However,
since there is usually only one accent mark to choose from, it has to cover
many different accents with the same letter, which can be a problem during
back-translation. In these cases, the "correct" opcode can help improve
back-translation of words that use these accented letters.
The "correct" opcodes is a characters-to-characters opcode both ways, so it
should be possible to implement the test part in the same manner in both
forward and backward translation.
The same goes for pass2-3-4, which are dots-to-dots opcodes both ways.
The problem is the "context" opcode, which is characters-to-dots one way
and
dots-to-characters the other way. So, it would have to work differently in
forward and backward translation. For back-translation, I would suggest
implementing the test part using the same implementation as for pass2-4.
Hope it makes sense.
Bue
-----Oprindelig meddelelse-----
Fra: liblouis-liblouisxml-bounce@xxxxxxxxxxxxx
[mailto:liblouis-liblouisxml-bounce@xxxxxxxxxxxxx] På vegne af Dave Mielke
Sendt: 11. januar 2017 06:04
Til: liblouis-liblouisxml@xxxxxxxxxxxxx
Emne: [liblouis-liblouisxml] Corrections pass during back translation.
Someone asked if the corrections pass during back translation makes sense
or
if it should be removed. My opinion is that it should be retained to deal
with cases where back translation gets it wrong.
--
Dave Mielke | 2213 Fox Crescent | The Bible is the very Word of
God.
Phone: 1-613-726-0014 | Ottawa, Ontario | http://Mielke.cc/bible/
EMail: Dave@xxxxxxxxx | Canada K2A 1H7 | http://FamilyRadio.org/
For a description of the software, to download it and links to project
pages
go to http://liblouis.org
For a description of the software, to download it and links to
project pages go to http://liblouis.org
