Hi Fae,
On 03-05-2018 16:18, Fæ wrote:
On 3 May 2018 at 19:54, Aidan Hogan
<ahogan(a)dcc.uchile.cl> wrote:
Hi all,
I am wondering what is the fastest/best way to get a local dump of English
Wikipedia in HTML? We are looking just for the current versions (no edit
history) of articles for the purposes of a research project.
We have been exploring using bliki [1] to do the conversion of the source
markup in the Wikipedia dumps to HTML, but the latest version seems to take
on average several seconds per article (including after the most common
templates have been downloaded and stored locally). This means it would take
several months to convert the dump.
We also considered using Nutch to crawl Wikipedia, but with a reasonable
crawl delay (5 seconds) it would several months to get a copy of every
article in HTML (or at least the "reachable" ones).
Hence we are a bit stuck right now and not sure how to proceed. Any help,
pointers or advice would be greatly appreciated!!
Best,
Aidan
[1]
https://bitbucket.org/axelclk/info.bliki.wiki/wiki/Home
Just in case you have not thought of it, how about taking the XML dump
and converting it to the format you are looking for?
Ref
https://en.wikipedia.org/wiki/Wikipedia:Database_download#English-language_…
Thanks for the pointer! We are currently attempting to do something like
that with bliki. The issue is that we are interested in the
semi-structured HTML elements (like lists, tables, etc.) which are often
generated through external templates with complex structures. Often from
the invocation of a template in an article, we cannot even tell if it
will generate a table, a list, a box, etc. E.g., it might say "Weather
box" in the markup, which gets converted to a table.
Although bliki can help us to interpret and expand those templates, each
page takes quite long, meaning months of computation time to get the
semi-structured data we want from the dump. Due to these templates, we
have not had much success yet with this route of taking the XML dump and
converting it to HTML (or even parsing it directly); hence we're still
looking for other options. :)
Cheers,
Aidan