Hi Bue, Aha! that explains a lot. I knew that libhyphen patterns are not the same as Tex patterns, but I was assuming you were using the correct format. The table at https://code.google.com/p/liblouis/source/browse/trunk/tables/hyph_da_DK.diclooks all right. The patgen program in combination with the Perl script seems like a good solution to me, or a re there any drawbacks? Do you expect to get better results from a "direct" implementation? Bert 2014/1/6 Bue Vester-Andersen <bue@xxxxxxxxxxxxxxxxxx> > Hello Bert, > > > > Thank you, but I don't think that would be an option in my case. Also, > that would mean that Liblouis wasn't really supporting Danish contraction > without the use of an additional hyphenation tool. That might be a show > stopper for Danish contraction in many situations where LibLouis is > currently being used successfully for making contracted Braille in other > languages. > > > > However, I think we have found the source of the problem. I currently have > a list of more than 33,000 manually hyphenated words. Until now, I have > been generating the hyphenation pattern file using the patgen program from > Tex live. But apparently, Liblouis needs another flavor of pattern file, > namely the type of pattern file used in OpenOffice and LibHyphen. I > thought the two were compatible, but apparently, they are not. I have just > read today that if you happen to use a Tex style pattern file in this > situation you will end up with a lot of bad hyphenations, which is exactly > what we are getting. It will simply "fail silently". > > > > I have found a Perl script that can convert a Tex style hyphenation file > into an OpenOffice style file. I am currently looking for a patgen program > that can generate the right type of hyphenation file streight away. > preferably one that can run under windows or be compiled to run under > Windows. > > If you have any suggestions, I would be very thankful in deed. > > > > Kind regards > > Bue > > > > > > *Fra:* liblouis-liblouisxml-bounce@xxxxxxxxxxxxx [mailto: > liblouis-liblouisxml-bounce@xxxxxxxxxxxxx] *På vegne af *Bert Frees > *Sendt:* 6. januar 2014 11:04 > *Til:* liblouis-liblouisxml@xxxxxxxxxxxxx > *Emne:* [liblouis-liblouisxml] Re: SV: Proposal for next liblouis release > date > > > > Hello Bue, > > > > Would it be an option for you to do the hyphenation beforehand, using a > standard tool? Liblouis has a function `translatePrehyphenated'. You give > it the input text and input hyphenation points, and it gives you the > translated text and output hyphenation points back. See > https://github.com/liblouis/liblouis/blob/master/liblouis/lou_translateString.c#L278. > We use this approach at SBS. > > > > Kind regards, > > Bert > > > > 2013/12/25 Bue Vester-Andersen <bue@xxxxxxxxxxxxxxxxxx> > > Hi Christian, > > I would very much like to get the hyphenation algorithm fixed before we > release the next version. At the moment, using the current Danish tables, > about half of the words from our test suite are hyphenated differently from > what would be expected with the given set of tables. The tables are now > excelent and produce very good hyphenation using standard hyphenation > tools. However, LibLouis does not seem to be able to interpret the > hyphenation tables correctly. So Danish contraction still looks very bad, > because it depends extensively on correct hyphenation. Other languages also > depend on the hyphenation. > > To me it looks like some loop counter getting out of synch or something > like it. Mesar is working hard on this issue, but any help or suggestions > would be greatly apreciated. We have a test suite of more than 30000 > manually hyphenated words. As I said, they hyphenate correctly using the > generated tables and standard hyphenation tools, but not with the LibLouis > hyphenator. If we don't get it fixed before this release, we may have to > wait yet another half year before we can have decent Danish contraction. > > Best regards and merry Christmas > Bue >