docs: add initial version

README.md

@@ -1,3 +1,32 @@
----
-license: cc-by-sa-4.0
----
+---
+license: cc-by-sa-4.0
+language:
+- bar
+pretty_name: Bavarian Wikipedia Dump
+size_categories:
+- 100K<n<1M
+---
+
+# Bavarian Wikipedia Dump
+
+This datasets hosts a recent Bavarian Wikipedia Dump that is used for various experiments within the [Bavarian NLP](https://huggingface.co/bavarian-nlp) organization.
+
+## Dataset Creation
+
+The latest dump was downloaded with:
+
+```bash
+wget https://dumps.wikimedia.org/barwiki/20250720/barwiki-20250720-pages-articles.xml.bz2
+```
+
+Then a patched version of [WikiExtractor](https://github.com/attardi/wikiextractor) was used (with Python 3.12.3 - newer versions were not working) to extract all articles into JSONL:
+
+```bash
+python3 -m wikiextractor.WikiExtractor --json --no-templates -o - barwiki-20250720-pages-articles.xml.bz2 > bar_wikipedia.jsonl
+```
+
+The final JSONL file was then uploaded here.
+
+## Stats
+
+The extracted JSONL file contains 43,917 articles.