Upload folder using huggingface_hub

README.md

@@ -23,7 +23,7 @@ dataset_info:
       description: Article title
     - name: text
       dtype: string
-      description: Article wikitext (main namespace only, no redirects)
+      description: Cleaned article text (templates removed, [[x|y]]→y, main ns, no redirects)
   splits:
     - name: train
       num_bytes: null
@@ -37,7 +37,7 @@ Hebrew Wikipedia article dump (main namespace only, redirects excluded), exporte
 ## Data
 
 - **Source**: [hewiki-20240901-pages-articles-multistream](https://dumps.wikimedia.org/hewiki/)
-- **Schema**: `id` (int64), `title` (string), `text` (string, wikitext)
+- **Schema**: `id` (int64), `title` (string), `text` (string, cleaned: templates removed, links simplified)
 - **License**: CC BY-SA 3.0 (Wikipedia)
 
 ## Usage
@@ -53,5 +53,4 @@ ds = load_dataset("ORG/REPO", trust_remote_code=True)
 
 ## Notes
 
-- Only main-namespace (ns=0) articles are included; talk pages, templates, etc. are excluded.
-- Redirects are excluded.
+- Only main-namespace (ns=0) articles; redirects excluded. Text is cleaned: `{{...}}` removed, `[[link|display]]`→display.

data/train.parquet

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:1fe603b67ad8b858508cb4fa6a67a4e5a05379f7ca9524f329f5ad87273b8300
-size 2324072965
+oid sha256:d98d237b129be8e752485c526a3e2a751f11c779a0ce32dfc8afcb0b29306f01
+size 1571369677