FCV data-use paper corpus (6 configs)

README.md

@@ -44,10 +44,6 @@ configs:
     path: "impact/validation.jsonl"
   - split: holdout
     path: "impact/holdout.jsonl"
-- config_name: synthetic
-  data_files:
-  - split: train
-    path: "synthetic/train.jsonl"
 - config_name: synthetic-relation
   data_files:
   - split: train
@@ -75,31 +71,26 @@ All splits are published verbatim exactly as used in the paper.
 
 | Config | Splits (records) |
 |---|---|
-| `extraction` | train: 5183, validation: 415, test: 1706 |
+| `extraction` | train: 22794, validation: 415, test: 1706 |
 | `relation` | train: 640, validation: 91, holdout: 21 |
 | `impact` | train: 604, validation: 151, holdout: 84 |
-| `synthetic` | train: 17611 |
 | `synthetic-relation` | train: 976 |
 | `sources` | train: 5445 |
 
 ## Schemas
 
-- `synthetic` / `sources`: `{"input", "output": {"entities":
+- `extraction` / `sources`: `{"input", "output": {"entities":
   {"named_data", "descriptive_data", "vague_data"}, "entity_descriptions"}}`.
-- `synthetic` merges all entity-extraction synthetic training data (flat-NER +
-  `json_structures` transformed to flat-NER) with a `source` column tagging
-  provenance (`ner_strict`, `positives_v14`, `step3_short_span`,
-  `structures_strict`, `approach_e_5000`, `old_training_synth`,
-  `old_data_synth`, `old_combined`, `manual_seeds`).
+- `extraction` rows carry a `meta` field tagging provenance: `FCV`
+  (forced-displacement / FCV corpus), `PRWP` (Policy Research Working Papers --
+  manually annotated + reviewed/validated), or `synthetic` (seed-based synthetic
+  training data, flat-NER). `train` = FCV + PRWP + synthetic; `validation` and
+  `test` are FCV-only.
 - `synthetic-relation`: relation-format synthetic training data (`{"input",
   "output": {"entities": {"named_data", "organization"}, "relations":
   [{"<type>": {"head", "tail"}}]}, "_meta"}`) from
   `relation_training_synthetic` and `relation_llm_generated`, with a `source`
   column.
-- `extraction` rows carry a `meta` field tagging provenance: `FCV`
-  (forced-displacement / FCV corpus) or `PRWP` (Policy Research Working Papers
-  -- manually annotated + reviewed/validated). `train` = FCV (2,473) + PRWP
-  (2,710); `validation` and `test` are FCV-only.
 - `relation`: `{"input", "output": {"entities": {"named_data",
   "organization"}, "relations": [{"<type>": {"head", "tail"}}]}, "_meta"}`
   with types `has_organization`, `used_by`, `has_acronym`, `has_timeframe`,

extraction/train.jsonl

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:246b973a130f61e95d909302eb62ca8af82b9b7cc58203202dfc95e90f1979d6
-size 10466229
+oid sha256:c4e4a1dcf4107a7558e571fb9bf50b62a3d9aa48822fe41a052b0e456ed6e298
+size 58100138