On Tuesday, November 4, 2014, Tino Didriksen <[email protected]> wrote:
> On 4 November 2014 08:41, Mikel L. Forcada <[email protected] > <javascript:_e(%7B%7D,'cvml','[email protected]');>> wrote: > >> For instance, I am amazed at how well you deal with formatting, links, >> etc. Is that code available? > > > This was explained / discussed in a previous thread: > - > http://www.mail-archive.com/apertium-stuff%40lists.sourceforge.net/msg04432.html > - http://www.mediawiki.org/wiki/Content_translation/LinearDoc > For mapping the HTML mark up from source content to translated content, we use an "upper casing algorithm" as detailed in https://www.mediawiki.org/wiki/Content_translation/Markup This is a post processing after MT. It works for the language pairs we currently support - es, ca, pt. But has lot of limitations as documented in the same page. We did some research on this and we have an alternate algorithm to do this mark up transfer from source content to translated content. See the section "Annotation mapping using translation subsequence approximation" at the end of that page. Sorry, we don't have a detailed write up other than that and the analysis document linked there. It is only a few days back we worked on this. This is something that can be ideally in apertium pipeline rather than a postprocessing tool outside apertium. Once we do a basic implementation and measure the results, we will discuss more about this to you people to bring this "HTML translation" support to apertium. Thanks Santhosh Thottingal
------------------------------------------------------------------------------
_______________________________________________ Apertium-stuff mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/apertium-stuff
