Upload ARETA gold reference dataset

README.md

@@ -5,12 +5,16 @@ configs:
 - config_name: annotations
   default: true
   data_files:
-  - split: test
+  - split: dev
     path: annotations.csv
-- config_name: contexts
+- config_name: contexts_nopnx
   data_files:
-  - split: test
-    path: contexts.csv
+  - split: dev
+    path: contexts_nopnx.csv
+- config_name: contexts_pnx
+  data_files:
+  - split: dev
+    path: contexts_pnx.csv
 ---
 
 # ArabiGEE Data
@@ -19,7 +23,7 @@ Paper title: **ArabiGEE: A Hierarchical Taxonomy for Arabic Grammatical Error Ex
 
 ## annotations.csv
 
-- `context_id`: Links the annotation row to `contexts.csv`.
+- `context_id`: Links the annotation row to either context file.
 - `pair_id`: Word-pair identifier in this dataset.
 - `original_pair_id`: Original pair identifier in the source data.
 - `error_id`: Error number within the same `pair_id`.
@@ -35,10 +39,20 @@ Paper title: **ArabiGEE: A Hierarchical Taxonomy for Arabic Grammatical Error Ex
 - `synt_code`: Syntactic explanation code.
 - `synt_explanation_text`: Syntactic explanation text.
 
-## contexts.csv
+## contexts_nopnx.csv
+
+- `context_id`: Context identifier in this dataset.
+- `source_dataset`: Source dataset label.
+- `source_split`: Source split. All rows use `dev`.
+- `source_sentence_id`: Original sentence identifier in the source data.
+- `src_context`: Erroneous sentence context without punctuation.
+- `target_context`: Corrected sentence context without punctuation.
+
+## contexts_pnx.csv
 
 - `context_id`: Context identifier in this dataset.
 - `source_dataset`: Source dataset label.
+- `source_split`: Source split. All rows use `dev`.
 - `source_sentence_id`: Original sentence identifier in the source data.
-- `src_context`: Erroneous sentence context.
-- `target_context`: Corrected sentence context.
+- `src_context`: Erroneous sentence context with punctuation.
+- `target_context`: Corrected sentence context with punctuation.