Dutch Medical Wikipedia

This repository documents how to download the dutch medical wikipedia.

1. Download all Dutch wikipedia articles

Dump file location: https://dumps.wikimedia.org/nlwiki

Wikipedia dump files are updated twice per month. Old versions are removed. It is recommended to use download the latest version, and include the file name in the download. For example, browse to 20200920/ and download nlwiki-20200920-pages-articles.xml.bz2 which is around 1.4GB.

There are also versions of the dump that are separated into separate files so that it is more suitable for parallel processing. This can cause some issues so it's recommended to use the complete -pages-articles.xml.bz2 file.

2. Download the medical categories

You can use the included categories_geneeskunde_depth4.txt file for this, or download the categories yourself using the following steps.

To select all medical Wikipedia articles, we use the top category for medical articles. In Dutch this is "Geneeskunde". However most dutch medical articles contain a more specific category, such as "Neurological condition" or "Cancer". Therefore, we use Wikipedia's Petscan to retrieve all categories that are related to the "Geneeskunde" category.

Go to https://petscan.wmflabs.org
Select language nl
Select depth 4. If you set depth too low, you will miss articles, but if it's set too high, you will include many false positives. After testing a few different categories, I think depth 4 is most accurate while having a good coverage. Note that Wikipedia categories are in a graph structure, not a tree structure.
Select the Page Properties tab, and make sure only "Categorie" Namespace is checked.
Press Do it!. This resulted in 841 categories on October 13, 2020.
Copy all categories to a .txt file. I did this by selecting the categories and copy/paste/reformat them using Atom and Excel. I saved it as categories_geneeskunde_depth4.txt and included it in this repository. Make sure the special characters are copied correctly. When I downloaded the categories directly as CSV, there were errors in copyting the special characters.

Copy all resulting category names to a text file, such as categories_geneeskunde_depth4. I did this by copy pasting the website data to Atom, to Excel, and then copy the correct values to a txt. When exporting to csv first, the special characters were copied incorrectly.

3. Download Wikiextractor

WikiExtractor is used extract all articles from the relevant categories. There were some issues with the latest stable release at the time of writing this document, and some fixes were necessary to correctly extract the categories, and make the code work with Dutch Wikipedia. Therefore, we use a fork of WikiExtractor in this project.

git clone -b stable_dutch --single-branch https://github.com/sandertan/wikiextractor/

TODO: Add fixes and customizable category to main repository.

4. Configure bash script to handle extracting and file naming

This is based on a Medium article.

Paths to set in extract_and_clean_wiki_dump.sh for your downloaded dump files and :

WIKI_DUMP_FILE_IN=nlwiki-20200901-pages-articles.xml.bz2
WIKI_DUMP_FILE_OUT=nlwiki_20200901_geneeskunde_depth4.txt
CATEGORIES_FILE=categories_geneeskunde_depth4.txt

5. Execute bash script

bash extract_and_clean_wiki_dump.sh

This process takes about 20 minutes using 4 cores.