On 16 May 2010 18:52, Jimmy O'Regan <[email protected]> wrote: > On 16 May 2010 17:46, Mikel L. Forcada <[email protected]> wrote: >> Dear Ian, >>> Mr. Forcada, >>> >> Please call me Mikel... >>> Thank you. I am not sure that I would not have the technical skills required >>> to initiate such a project, but I have fairly good organizational and >>> linguistic skills to contribute to coordination of such a project. >> That's excellent news. > > You don't really need a whole lot of technical skill; as long as > you're willing to contribute, there will usually be somebody willing > to help you to contribute. > >>> My brother >>> Mark Shaw is also a Linux expert and has developed his own Linux >>> distribution. >>> >> Which one? >>> He could help in this project and has fairly extensive contacts throughout >>> the >>> world with programmers. >> Our project has also capable programmers and linguists that could help. >> I am indeed sending a copy of this message to our mailing list, >> [email protected], so that the Apertium community is informed >> of your intentions and to see if someone out there is also able to lend >> a hand. >>> What is your protocol in developing language pairing? >>> >> There are many ways in which languages are paired, depending on the data >> available. In the case of en-ia or es-ia, we already have rather >> extensive es and en dictionaries. Structural transfer rules for es-ia >> would be rather easy to write, and those for en-ia could be based in >> those we already have for en-ca or en-es. One would have to build >> complete monolingual dictionaries for Interlingua, and also bilingual >> es-ia or en-ia dictionaries. If you wanted translation from ia, then one >> would have to train a part-of-speech tagger for ia. >> >>> How would we be able to work with you to put this on the Apertium platform? >> I can sign you up as developers in our Sourceforge platform, and you can >> start working on the language pair there. I think initially the new >> pairs go to some kind of incubator and then to the trunk. Can anyone on >> the list help us on what would be the protocol? >> >> Tu pote leger anque >> http://wiki.apertium.org/wiki/Apertium_New_Language_Pair_HOWTO >> > > The usual IRC experience is: > > * Make contact (done) > * Read and adapt the New Language Pair HOWTO to the new language pair > * Ask any questions you may have > * Send the work you did on the new language pair to an existing > developer; that developer then adds it to the incubator and gives you > the URL to the language pair (mostly, that's because Apertium's SVN > repository is huge and nobody wants to expect anyone to check out the > full ~4Gb) > * You are nominated as a developer; if the nomination is seconded > you're added (this is a relatively new rule; nobody has ever not been > seconded) > > Usually, someone will know some source of data for the language pair > and/or pitch in to help you get under way. > > Language pairs are moved to trunk either after they have been > released, or if there is a funded group working on them (more or less, > if the disappearance of a single person won't be the end of the pair). > > > I would recommend es-ia first; from Mikel's writing, it looks close > enough to Spanish that apertium-transfer-tools could be used (given an > analyser, a bilingual lexicon, and a small corpus); similar languages > also yield better translations. Converting en-es rules is something an > experienced user can do in a matter of days, but you need to get the > experience first (which you could, by editing the generated rules) - > otherwise, it may become overwhelming quite quickly.
I started to put together a rudimentary es-ia translator: $ echo "veo el gato" |apertium -d . test-es-ia Io vide #le catto As you can see (by the # mark), there's a small problem straight away - there are many more where that came from. In any case, it's a starting point. You can check it out from SVN: svn co https://apertium.svn.sourceforge.net/svnroot/incubator/apertium-es-ia/ apertium-es-ia -- <Leftmost> jimregan, that's because deep inside you, you are evil. <Leftmost> Also not-so-deep inside you. ------------------------------------------------------------------------------ _______________________________________________ Apertium-stuff mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/apertium-stuff
