16 comments

[ 3.1 ms ] story [ 37.4 ms ] thread
What about DBPedia http://dbpedia.org/page/The_Shining_(film) ?

That's available as JSON, and much larger community/effort behind it.

There's a much larger community/effort behind DBPedia than Wikipedia?

Or are you comparing DBPedia to this utility that converts Wikipedia data dumps?

Edit: looks like I need to do some more research into DBPedia. It sounds like that community has created a dataset that categorizes Wikipedia article (e.g. "things" and "facts")

Correct me if I am wrong, but I do not think that DBPedia includes the full Wikipedia articles (only extended abstracts). So I believe that json dumps of DBPedia and Wikipedia dumps produced using this tool are two different things.
In my understanding, DBpedia is a project for extracting data mainly from Wikipedia infoboxes (not whole data dumps) by collaboratively creating rules for converting data into cleaner schema that enables to perform a query such as SPARQL. I think this is a project that directly converts wikipedia dump xml to JSON for easier manipulation, which differs from DBpedia.
Hi, I'm the author of Json-Wikipedia. It's different from DB-Pedia. DBPedia converts the semistructured data containined in the infoboxes in rdf triples, while Json-Wikipedia converts ALL the content of the page (text, infoboxes, links) into a JSON object. I did this because I need some features that were not available on dbpedia (e.g., the whole text of an article, all the anchors referring to an article etc etc). Also Json-Wikipedia simplifies the processing of a Wikipedia Dump on Hadoop.
So, does it parse the markdown of the text as well? I could not quite get it from the documentation. It would be quite exciting if it could - the markdown is notoriously complex, and creates a lot of problem if all you want is te body of text (e.g. for training a machine learning algorithm).
I doubt it (though I haven't tested it). Wikipedia's API though lets you get an HTML version of a page.
Wikipedia articles are not written in markdown but in wikicode. Wikicode is much more complex to parse, but allows more complex formatting. You may be looking for the word 'markup' which is a general term that included markdown, HTML, wikicode, bbcode etc.

I doubt this library parses the wikicode. What format would it parse it into? As matthewarkin said, there is a Wikipedia API to convert wikicode into HTML.

It does that indeed! It will process a field paragraph, containing a list of parsed paragraphs in the article, as simple UTF8 text. Lists and Tables are collected in separated fields, have a look to the Article object. For the wikicode parsing I relied on the JWPL Library http://www.ukp.tu-darmstadt.de/software/jwpl/
Why would I need the dump in JSON? XML is for sure uglier but at least there are efficient SAX parsers.
I would like to know this as well. What is gained by using this tool... converting to JSON?
JSON is cool and hip.

More seriously, it has been gaining popularity because of it's direct relationship to built-in structures in most programming languages.

For example: XML doesn't have a concept of arrays, but accomplishes storing collections in it's own special way. It's up to the parser to determine how this should be referenced and interacted-with, while with JSON there's no question: it's an array.

For example, now I know for a fact I can easily manipulate Wikipedia's data dump using Python thanks to this JSON change.

    import json

    js = json.loads(wikidump)
That's it. The only reason it's so simple is thanks to this JSON conversion.

XML? No clue how to manipulate it properly. I could figure it out of course, but now I don't have to. JSON + Python just works™.