Update README.md

README.md

@@ -89,4 +89,32 @@ configs:
         path: train400k_val2k_test2k_sentence/val.csv
       - split: test
         path: train400k_val2k_test2k_sentence/test.csv
----
+---
+# WikiEditBias Dataset
+
+Wikipedia Editorial Bias Dataset. This dataset serves for the task of detecting biases in Wikipedia historical revisions. This dataset is generated by tracking Wikipedia revisions and corresponding editors' bias labels from the [MediaWiki Historical Dump](https://dumps.wikimedia.org/other/mediawiki_history/readme.html).
+
+## Uses
+
+### Direct Use
+
+<!-- This section describes suitable use cases for the dataset. -->
+
+```
+from datasets import load_dataset
+dataset = load_dataset("fgs218ok/WikiEditBias")
+```
+
+## Dataset Structure
+
+<!-- This section provides a description of the dataset fields, and additional information about the dataset structure such as criteria used to create the splits, relationships between data points, etc. -->
+The WikiEditBias Dataset has two data formats: 
+* Edit diff format: Contains by sentence pairs extracted from sentence-level differences of Wikipedia revisions. For each .csv file there are 3 fields:
+  * label: 0 refers to the non-biased/neutral edits and 1 refers to the biased edits.
+  * old_text: pre-edit sentence-level texts
+  * new_text: after-edit sentence-level texts
+* Sentence format: Contains sentences extracted from the Wikipedia revisions. The fields are similar to the edit diff format:
+  * label: 0 refers to the non-biased/neutral edits and 1 refers to the biased edits.
+  * text: sentence-level texts of edits.
+  
+For each format there are five scales of data given: 10k, 50k, 100k, 200k, 400k