[Corpora-List] Corpus from Blogs required.
Trilok Khairnar
trilokgk at gmail.com
Mon Apr 4 08:45:44 UTC 2005
Hello Jean-Phi, Gilad
Thanks for the inputs.
Permalinks and Technorati APIs will definitely be useful.
Technorati APIS provide - inbound and outbound links of a blog, basic
user and blog info etc. but not the list of posts on a blog and their
text.
On the other hand, permalinks should be useful to extract the text of
one blog post at a time though surrounding text on the blog like
badges and blogroll will be included too. (Looks like a hack will be
required to extract only the text of a post when permalink is
available.)
I will try this sometime using Atom.Net and RSS.Net libraries and let
the list-members know.
Thanks,
Trilok.
On Mar 31, 2005 4:05 AM, Jean-Phi <jpprost at gmail.com> wrote:
> Hi,
>
> > In the absence of such corpus and APIs, I am thinking of doing this by
> > 1] using RSS, ATOM feed parsers on some OPML files to get URLs for blog posts
> > 2] Extracting the text (easier if the blog template format is known)
>
> It might not be that easy: I suspect that many blogs use some sort of
> Content Management System, which basically means that the texts are
> stored in a database, and are only presented in the blog dynamically,
> on request.In such cases my guess is that you'll probably need to know
> a minimum about the database structure in order to query it --unless,
> of course, the site provides you with an RSS feed. Or do I miss
> something?
>
> Some blog host sites may sometimes also couple the dynamic rendering
> with a permanent html link for each text. http://www.blogger.com/ (now
> owned by google) does provide both these features: RSS feed and
> permanent link. I don't hold any shares, though...
>
> Cheers,
> --
> Jean-Philippe Prost
> Centre for Language Technology
> Macquarie University ~ Sydney, Australia
> and
> Laboratoire Parole et Langage (Speech & Language Lab.)
> Université de Provence ~ Aix-en-Provence, France
> <http://www.ics.mq.edu.au/~jpprost/>
> _______________________________________________
>
>
More information about the Corpora
mailing list