FCV data-use paper corpus (6 configs)

.gitattributes

@@ -60,3 +60,5 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *.webm filter=lfs diff=lfs merge=lfs -text
 documents/all.jsonl filter=lfs diff=lfs merge=lfs -text
 sources/all.jsonl filter=lfs diff=lfs merge=lfs -text
+extraction/train.jsonl filter=lfs diff=lfs merge=lfs -text
+sources/train.jsonl filter=lfs diff=lfs merge=lfs -text

README.md

@@ -50,8 +50,8 @@ configs:
     path: "synthetic/train.jsonl"
 - config_name: sources
   data_files:
-  - split: all
-    path: "sources/all.jsonl"
+  - split: train
+    path: "sources/train.jsonl"
 ---
 
 # Dataset Card for FCV Data-Use Paper
@@ -75,7 +75,7 @@ All splits are published verbatim exactly as used in the paper.
 | `relation` | train: 640, validation: 91, holdout: 21 |
 | `impact` | train: 604, validation: 151, holdout: 84 |
 | `synthetic` | train: 1794 |
-| `sources` | all: 5445 |
+| `sources` | train: 5445 |
 
 ## Schemas
 

sources/train.jsonl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:565899f1ce3e9c38b357a3dd153054d70f08f0f2744088ba1b70332f700057f1
+size 12763660