[lojban] Re: Grammar checking wikipedia bot.

[Pierre Abbat <lojban-out@lojban.org>]

On Saturday 26 August 2006 21:09, Einar Faanes wrote:
> coi ro do
>
> I mentioned this on the irc-channel earlier today, but I think I should
> post it here as well. I have an idea which I think is possible and which
> I think should be set alive. When lojban is parseable (and we have a
> parser) we should take advantage of that by automatically check the
> lojban wikipedia for spelling errors by channeling the text through
> jbofi'e. I think that this can be done by using the wikimedia bot
> framework, which is among other things used to update and add
> interlanguage links.
>
> It should be possible to make the bot download a page, strip it of
> non-lojban elements (wikimarkup etc.), check the text and post an
> errormessage on either the articles discussion page or a centralized
> reference page. The bot is written in python. I'm no programmer, but
> know others here which are and which may find this interesting.

Sounds good, though there are a couple of things that would cause false 
errors:
1. Some words are valid fu'ivla, but jbofi'e doesn't recognize them, such as 
{srutio} (a discarded form for {strutione}, ostrich) and {largectremia} 
(crape myrtle). Also the PEG accepts some lujvo made with fu'ivla that 
jbofi'e doesn't.
2. Some tables don't parse if you just remove the markup. The prefix chart in 
[[treci'e]] is set up so that you put the cells in the row in the blanks in 
the sentence formed by the header row. If you removed the markup in a table 
in the English Wikipedia, the result wouldn't parse in English either.

phma


To unsubscribe from this list, send mail to lojban-list-request@lojban.org
with the subject unsubscribe, or go to http://www.lojban.org/lsg2/, or if
you're really stuck, send mail to secretary@lojban.org for help.