A completed run of Analysis.rmd can be found at
http://darrenabramson.github.io/CitationNeeded/Analysis.html. The results show pretty clearly that unless some coordinated top-down plan is initiated, citation needed tagged edits tend to just remain in Wikipedia. I'm not sure this is a bad thing.

The purpose of this code is to provide an example of a behavioral analysis of Wikipedia. The intended application is for providing empirical justification for a controversial epistemological category.

You'll need bzip2.jar and wikixmlj-r43.jar extracted to where you compile.

On my computer the following worked:

javac Main.java

java Main "WikipediaXML_Compressed.bz2"

At the command line, this should give you three numbers along with other possible output:

1. Number of pages in the XML dump with a {{citation needed}} tag.

2. Total number of pages in the XML dump.

3. Total number of {{citation needed}} tags in the XML dump.

4. Any malformed 'citation needed' items. At present, there are only two, both of which are found on https://en.wikipedia.org/wiki/Wikipedia:Biographies_of_living_persons/Risk_lists/List_2 in the form of partially quoted citation needed tags.

5. Any tags that appear on a page prior to PRECEDING_SENTENCE_LENGTH characters from the beginning. Note that this includes markup characters. Each such page, with the character number of the first appearance of a citation needed tag, is listed in the output.

To pipe the output to a log for a run enter something like the following:

java Main "enwiki-pages-articles.xml.bz2" -option | tee enwiki-pages-articles-DateTime.log

It will also create a file with a name like the following:

2015.07.21 AD at 13:42:17 ADT.WikiDataFileName.bz2.csv

At present, this is a table with three columns:

• pageName
• tagContents
• precedingSentence

pageName is the name of a Wikipedia page that contains a citation needed tag. tagcontents is the information contained in the citation needed tag. In virtually every case this contains date information for when the tag was created. precedingSentence is, rather arbitrarily, the 30 characters preceding the tag.

In R studio, the following code imports the csv file into a table with the correct dimensions, matching the terminal output for number of tags (excel seems to mangle some of the delimiters).

cnTable <- test read.csv("2015.07.21 AD at 13:42:17 ADT.WikiDataFileName.bz2.csv", header=TRUE, quote="")

It may be preferable to add a stringsAsFactors=FALSE argument.

The next important step is to strip out and import date information. I think there's already enough information to get halflife from a single dump.