From: Justin Bailey <jgbailey@...>
Date: 2006-07-06T02:59:38+09:00
Subject: Re: [SOC] progress on "Automated Wrapper Generation for Information Extraction"

------=_Part_3611_16316182.1152122376082
Content-Type: text/plain; charset=ISO-8859-1; format=flowed
Content-Transfer-Encoding: 7bit
Content-Disposition: inline

On 7/5/06, A. S. Bradbury <asbradbury@tekcentral.org> wrote:
>
>
> I think I understand what you're asking here. Given a page such as
> http://www.blueletterbible.org/Gen/Gen001.html - you might define the
> structure such as:
>   doc_tree = Ariel::StructureNode.new do |r|
>     r.verse_list do |v|
>       v.reference
>       v.text
>       end
>     end
>   end
>
> After training, given a chapter from that site, the generated rules could
> be
> applied to produce a list of reference and verse pairs. So you could then
> just search that list for the member of verse_list with reference=="Gen
> 1:1".
> Is that basically what you're thinking? If so, Ariel should be very suited
> to
> that task. That actually looks like quite a nice page to use in testing...


Yes, that's even more than I expected. I especially like how it rolls up the
verses.

Yes, I'm familiar with microformats. As Austin says, perhaps they're not
> right
> for my project as it takes an approach that should be capable of dealing
> with
> a wide variety of semi-structured documents. Is being intrusive a problem?


If you did have a "microformats" (or similary non-intrusive means) of
marking up a page, it could allow anyone to provid an "ariel" interface to
their data right in the page. Now, I don't know if anyone would do that -
but it sounds cool to just point Ariel at a page/site that is compliant and
have basically at a data-access layer for it. Pretty blue-sky though.

Justin

------=_Part_3611_16316182.1152122376082--