Record visual conversion provenance and exclusions

provenance.json

@@ -0,0 +1,31 @@
+{
+  "context_construction": "one \"Speaker, Speaker: utterance\" line per native utterance, joined by newline; answers relocated from utterance_id + inner token span (or speaker name) to a character offset",
+  "conversion_sha256": "578eedfadd2cde779580e526d0aed05f766a85f40b206ce1d8d4158ebed6062c",
+  "dropped_answer_spans": {
+    "train": 2
+  },
+  "excluded_count": 0,
+  "license": "apache-2.0",
+  "license_evidence": "Repository LICENSE.txt: Apache License, Version 2.0.",
+  "mirror": "tasksource/friendsqa",
+  "packaging_script": "scripts/repackage_dataset/dialogue.py:friendsqa",
+  "revision": "9ce6c98a6acf13c87d96b6811e4409b5de43bdfe",
+  "source": "https://github.com/emorynlp/FriendsQA",
+  "source_hashes": {
+    "friendsqa_dev.json": "3a85d3fcb175589ffd050fb4bf431da77a6e12589c8ec03f1d54060ce33ec703",
+    "friendsqa_trn.json": "d63685693b30cee5d33b7cf2e6408cd284c05e49be1b3b06954c4d01cfe1c91b",
+    "friendsqa_tst.json": "a98f24d4b04ac9f8d9468088febf304bd90eada134537277c1a425e31c0372bc"
+  },
+  "split_policy": "native trn/dev/tst partitions mapped to train/validation/test",
+  "splits": {
+    "test": {
+      "rows": 1201
+    },
+    "train": {
+      "rows": 9874
+    },
+    "validation": {
+      "rows": 1182
+    }
+  }
+}