Upload folder using huggingface_hub
Browse files- README.md +63 -0
- articles.parquet +3 -0
- gold.parquet +3 -0
- split.json +289 -0
- unions.json +0 -0
README.md
ADDED
|
@@ -0,0 +1,63 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
license: cc-by-4.0
|
| 3 |
+
task_categories:
|
| 4 |
+
- text-classification
|
| 5 |
+
- token-classification
|
| 6 |
+
language:
|
| 7 |
+
- es
|
| 8 |
+
tags:
|
| 9 |
+
- signed-graphs
|
| 10 |
+
- relation-extraction
|
| 11 |
+
- political-networks
|
| 12 |
+
- text2graph
|
| 13 |
+
- benchmark
|
| 14 |
+
pretty_name: text2signed-graph Gold Standard (Synthetic v2)
|
| 15 |
+
size_categories:
|
| 16 |
+
- n<1K
|
| 17 |
+
---
|
| 18 |
+
|
| 19 |
+
# text2signed-graph — Gold Standard (Synthetic v2)
|
| 20 |
+
|
| 21 |
+
A synthetic benchmark for **extracting signed political relation graphs from news
|
| 22 |
+
articles**. Each relation is a directed, signed triple between political actors:
|
| 23 |
+
`(actor_u → act_type → actor_v, polarity)`, grounded in an evidence quote.
|
| 24 |
+
|
| 25 |
+
Built to evaluate `text2SG` extraction pipelines (NER → entity resolution → relation
|
| 26 |
+
extraction) in a **precision-first** setting, with planted distractors to penalize
|
| 27 |
+
over-extraction.
|
| 28 |
+
|
| 29 |
+
## Contents
|
| 30 |
+
|
| 31 |
+
| File | Rows | Description |
|
| 32 |
+
|---|---|---|
|
| 33 |
+
| `articles.parquet` | 287 | synthetic news articles (incl. distractors) |
|
| 34 |
+
| `gold.parquet` | 914 | gold signed relations |
|
| 35 |
+
| `split.json` | — | `train` (207) / `test` (75) article split + seed |
|
| 36 |
+
| `unions.json` | — | adjudicated entity unions (alias → canonical actor) |
|
| 37 |
+
|
| 38 |
+
### `gold.parquet` schema
|
| 39 |
+
`article_id`, `u_from`, `u_to` (actor ids), `act_type` (e.g. `endorses`, `attacks`,
|
| 40 |
+
`allies_with`, `questions`…), **`polarity`** (`positive` / `negative` / `neutral` — the
|
| 41 |
+
signed edge), `is_reactive`, `issue`, `evidence_quote`, `source` (`planted` / …),
|
| 42 |
+
`n_inclusion_votes`, `dispute_type`.
|
| 43 |
+
|
| 44 |
+
### `articles.parquet` schema
|
| 45 |
+
`article_id`, `title`, `body`, `dominio`, `registro`, `medio`, `dureza`,
|
| 46 |
+
**`es_distractor`** (distractor flag), `writer`, `difficulty`.
|
| 47 |
+
|
| 48 |
+
## Why synthetic
|
| 49 |
+
|
| 50 |
+
Real Chilean political news carry copyright (CC-BY-NC). This synthetic gold reproduces the
|
| 51 |
+
register, difficulty, and structure of real coverage — including **distractor articles and
|
| 52 |
+
near-miss relations** — so it can be shared openly as a reproducible benchmark.
|
| 53 |
+
|
| 54 |
+
## Evaluation
|
| 55 |
+
|
| 56 |
+
Score a predicted relation set against `gold.parquet` on the `test` split. The reference
|
| 57 |
+
extractor (`id15` / `gemini_champion_v2`) reaches **f0.5 ≈ 0.89** on detection (undirected),
|
| 58 |
+
precision-first. Distractors and `dispute_type` test robustness to over-extraction.
|
| 59 |
+
|
| 60 |
+
## Citation
|
| 61 |
+
|
| 62 |
+
Part of the **chilean-political-dataset / text2SG** line of work on signed political
|
| 63 |
+
networks (2014–2026). Built by Benjamín Palacios.
|
articles.parquet
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:33ce3c5a18495cb1ee9adc1c6923fa0613d4a244e8f6f60f9c736ba6140456d8
|
| 3 |
+
size 358504
|
gold.parquet
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:1551d95a2176355e4387787a23e12cd141e3b33bed76c97e178514aa2c48c16b
|
| 3 |
+
size 29721
|
split.json
ADDED
|
@@ -0,0 +1,289 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"train": [
|
| 3 |
+
"SV1-001",
|
| 4 |
+
"SV1-002",
|
| 5 |
+
"SV1-004",
|
| 6 |
+
"SV1-005",
|
| 7 |
+
"SV1-006",
|
| 8 |
+
"SV1-007",
|
| 9 |
+
"SV1-008",
|
| 10 |
+
"SV1-010",
|
| 11 |
+
"SV1-012",
|
| 12 |
+
"SV1-013",
|
| 13 |
+
"SV1-015",
|
| 14 |
+
"SV1-016",
|
| 15 |
+
"SV1-017",
|
| 16 |
+
"SV1-018",
|
| 17 |
+
"SV1-021",
|
| 18 |
+
"SV1-022",
|
| 19 |
+
"SV1-023",
|
| 20 |
+
"SV1-024",
|
| 21 |
+
"SV1-025",
|
| 22 |
+
"SV1-026",
|
| 23 |
+
"SV1-027",
|
| 24 |
+
"SV1-028",
|
| 25 |
+
"SV1-029",
|
| 26 |
+
"SV1-031",
|
| 27 |
+
"SV1-034",
|
| 28 |
+
"SV1-035",
|
| 29 |
+
"SV1-037",
|
| 30 |
+
"SV1-039",
|
| 31 |
+
"SV1-040",
|
| 32 |
+
"SV1-042",
|
| 33 |
+
"SV1-043",
|
| 34 |
+
"SV1-044",
|
| 35 |
+
"SV1-047",
|
| 36 |
+
"SV1-048",
|
| 37 |
+
"SV1-051",
|
| 38 |
+
"SV1-052",
|
| 39 |
+
"SV1-053",
|
| 40 |
+
"SV1-058",
|
| 41 |
+
"SV1-059",
|
| 42 |
+
"SV1-060",
|
| 43 |
+
"SV1-062",
|
| 44 |
+
"SV1-063",
|
| 45 |
+
"SV1-064",
|
| 46 |
+
"SV1-065",
|
| 47 |
+
"SV1-066",
|
| 48 |
+
"SV1-067",
|
| 49 |
+
"SV1-069",
|
| 50 |
+
"SV1-070",
|
| 51 |
+
"SV1-071",
|
| 52 |
+
"SV1-072",
|
| 53 |
+
"SV1-073",
|
| 54 |
+
"SV1-075",
|
| 55 |
+
"SV1-076",
|
| 56 |
+
"SV1-077",
|
| 57 |
+
"SV1-078",
|
| 58 |
+
"SV1-079",
|
| 59 |
+
"SV1-080",
|
| 60 |
+
"SV1-081",
|
| 61 |
+
"SV1-082",
|
| 62 |
+
"SV1-083",
|
| 63 |
+
"SV1-084",
|
| 64 |
+
"SV1-085",
|
| 65 |
+
"SV1-086",
|
| 66 |
+
"SV1-087",
|
| 67 |
+
"SV1-088",
|
| 68 |
+
"SV1-089",
|
| 69 |
+
"SV1-091",
|
| 70 |
+
"SV1-092",
|
| 71 |
+
"SV1-093",
|
| 72 |
+
"SV1-094",
|
| 73 |
+
"SV1-095",
|
| 74 |
+
"SV1-096",
|
| 75 |
+
"SV1-097",
|
| 76 |
+
"SV1-098",
|
| 77 |
+
"SV1-101",
|
| 78 |
+
"SV1-103",
|
| 79 |
+
"SV1-104",
|
| 80 |
+
"SV1-105",
|
| 81 |
+
"SV1-106",
|
| 82 |
+
"SV1-107",
|
| 83 |
+
"SV1-108",
|
| 84 |
+
"SV1-109",
|
| 85 |
+
"SV1-110",
|
| 86 |
+
"SV1-113",
|
| 87 |
+
"SV1-114",
|
| 88 |
+
"SV1-115",
|
| 89 |
+
"SV1-117",
|
| 90 |
+
"SV1-120",
|
| 91 |
+
"SV1-123",
|
| 92 |
+
"SV1-125",
|
| 93 |
+
"SV1-128",
|
| 94 |
+
"SV1-129",
|
| 95 |
+
"SV1-130",
|
| 96 |
+
"SV1-131",
|
| 97 |
+
"SV1-134",
|
| 98 |
+
"SV1-135",
|
| 99 |
+
"SV1-136",
|
| 100 |
+
"SV1-138",
|
| 101 |
+
"SV1-139",
|
| 102 |
+
"SV1-141",
|
| 103 |
+
"SV1-142",
|
| 104 |
+
"SV1-144",
|
| 105 |
+
"SV1-145",
|
| 106 |
+
"SV1-146",
|
| 107 |
+
"SV1-148",
|
| 108 |
+
"SV1-149",
|
| 109 |
+
"SV1-150",
|
| 110 |
+
"SV1-152",
|
| 111 |
+
"SV1-156",
|
| 112 |
+
"SV1-157",
|
| 113 |
+
"SV1-158",
|
| 114 |
+
"SV1-159",
|
| 115 |
+
"SV1-160",
|
| 116 |
+
"SV1-161",
|
| 117 |
+
"SV1-163",
|
| 118 |
+
"SV1-164",
|
| 119 |
+
"SV1-165",
|
| 120 |
+
"SV1-166",
|
| 121 |
+
"SV1-167",
|
| 122 |
+
"SV1-168",
|
| 123 |
+
"SV1-169",
|
| 124 |
+
"SV1-170",
|
| 125 |
+
"SV1-171",
|
| 126 |
+
"SV1-172",
|
| 127 |
+
"SV1-173",
|
| 128 |
+
"SV1-174",
|
| 129 |
+
"SV1-175",
|
| 130 |
+
"SV1-176",
|
| 131 |
+
"SV1-179",
|
| 132 |
+
"SV1-180",
|
| 133 |
+
"SV1-181",
|
| 134 |
+
"SV1-182",
|
| 135 |
+
"SV1-183",
|
| 136 |
+
"SV1-184",
|
| 137 |
+
"SV1-186",
|
| 138 |
+
"SV1-187",
|
| 139 |
+
"SV1-188",
|
| 140 |
+
"SV1-190",
|
| 141 |
+
"SV1-191",
|
| 142 |
+
"SV1-192",
|
| 143 |
+
"SV1-193",
|
| 144 |
+
"SV1-194",
|
| 145 |
+
"SV1-196",
|
| 146 |
+
"SV1-197",
|
| 147 |
+
"SV1-199",
|
| 148 |
+
"SV1-200",
|
| 149 |
+
"gen_003",
|
| 150 |
+
"gen_004",
|
| 151 |
+
"gen_006",
|
| 152 |
+
"gen_007",
|
| 153 |
+
"gen_008",
|
| 154 |
+
"gen_015",
|
| 155 |
+
"gen_016",
|
| 156 |
+
"gen_017",
|
| 157 |
+
"gen_018",
|
| 158 |
+
"gen_019",
|
| 159 |
+
"gen_021",
|
| 160 |
+
"gen_023",
|
| 161 |
+
"gen_026",
|
| 162 |
+
"gen_027",
|
| 163 |
+
"gen_031",
|
| 164 |
+
"gen_032",
|
| 165 |
+
"gen_033",
|
| 166 |
+
"gen_034",
|
| 167 |
+
"gen_035",
|
| 168 |
+
"gen_039",
|
| 169 |
+
"gen_040",
|
| 170 |
+
"gen_041",
|
| 171 |
+
"gen_043",
|
| 172 |
+
"gen_044",
|
| 173 |
+
"gen_045",
|
| 174 |
+
"gen_046",
|
| 175 |
+
"gen_047",
|
| 176 |
+
"gen_048",
|
| 177 |
+
"gen_050",
|
| 178 |
+
"gen_052",
|
| 179 |
+
"gen_053",
|
| 180 |
+
"gen_054",
|
| 181 |
+
"gen_056",
|
| 182 |
+
"gen_058",
|
| 183 |
+
"gen_060",
|
| 184 |
+
"gen_061",
|
| 185 |
+
"gen_062",
|
| 186 |
+
"gen_063",
|
| 187 |
+
"gen_064",
|
| 188 |
+
"gen_065",
|
| 189 |
+
"gen_066",
|
| 190 |
+
"gen_067",
|
| 191 |
+
"gen_068",
|
| 192 |
+
"gen_070",
|
| 193 |
+
"gen_072",
|
| 194 |
+
"gen_073",
|
| 195 |
+
"gen_075",
|
| 196 |
+
"gen_076",
|
| 197 |
+
"gen_078",
|
| 198 |
+
"gen_079",
|
| 199 |
+
"gen_081",
|
| 200 |
+
"gen_083",
|
| 201 |
+
"gen_084",
|
| 202 |
+
"gen_085",
|
| 203 |
+
"gen_086",
|
| 204 |
+
"gen_087",
|
| 205 |
+
"gen_089",
|
| 206 |
+
"gen_090",
|
| 207 |
+
"gen_092",
|
| 208 |
+
"gen_093",
|
| 209 |
+
"gen_094"
|
| 210 |
+
],
|
| 211 |
+
"test": [
|
| 212 |
+
"SV1-003",
|
| 213 |
+
"SV1-009",
|
| 214 |
+
"SV1-011",
|
| 215 |
+
"SV1-014",
|
| 216 |
+
"SV1-019",
|
| 217 |
+
"SV1-030",
|
| 218 |
+
"SV1-032",
|
| 219 |
+
"SV1-033",
|
| 220 |
+
"SV1-036",
|
| 221 |
+
"SV1-038",
|
| 222 |
+
"SV1-041",
|
| 223 |
+
"SV1-046",
|
| 224 |
+
"SV1-049",
|
| 225 |
+
"SV1-050",
|
| 226 |
+
"SV1-054",
|
| 227 |
+
"SV1-055",
|
| 228 |
+
"SV1-056",
|
| 229 |
+
"SV1-057",
|
| 230 |
+
"SV1-061",
|
| 231 |
+
"SV1-068",
|
| 232 |
+
"SV1-074",
|
| 233 |
+
"SV1-090",
|
| 234 |
+
"SV1-099",
|
| 235 |
+
"SV1-100",
|
| 236 |
+
"SV1-102",
|
| 237 |
+
"SV1-116",
|
| 238 |
+
"SV1-118",
|
| 239 |
+
"SV1-121",
|
| 240 |
+
"SV1-122",
|
| 241 |
+
"SV1-124",
|
| 242 |
+
"SV1-126",
|
| 243 |
+
"SV1-127",
|
| 244 |
+
"SV1-132",
|
| 245 |
+
"SV1-133",
|
| 246 |
+
"SV1-137",
|
| 247 |
+
"SV1-140",
|
| 248 |
+
"SV1-143",
|
| 249 |
+
"SV1-147",
|
| 250 |
+
"SV1-151",
|
| 251 |
+
"SV1-153",
|
| 252 |
+
"SV1-154",
|
| 253 |
+
"SV1-155",
|
| 254 |
+
"SV1-162",
|
| 255 |
+
"SV1-177",
|
| 256 |
+
"SV1-178",
|
| 257 |
+
"SV1-185",
|
| 258 |
+
"SV1-189",
|
| 259 |
+
"SV1-195",
|
| 260 |
+
"SV1-198",
|
| 261 |
+
"gen_001",
|
| 262 |
+
"gen_002",
|
| 263 |
+
"gen_005",
|
| 264 |
+
"gen_009",
|
| 265 |
+
"gen_010",
|
| 266 |
+
"gen_011",
|
| 267 |
+
"gen_012",
|
| 268 |
+
"gen_013",
|
| 269 |
+
"gen_014",
|
| 270 |
+
"gen_020",
|
| 271 |
+
"gen_022",
|
| 272 |
+
"gen_029",
|
| 273 |
+
"gen_042",
|
| 274 |
+
"gen_049",
|
| 275 |
+
"gen_051",
|
| 276 |
+
"gen_055",
|
| 277 |
+
"gen_057",
|
| 278 |
+
"gen_059",
|
| 279 |
+
"gen_069",
|
| 280 |
+
"gen_071",
|
| 281 |
+
"gen_074",
|
| 282 |
+
"gen_077",
|
| 283 |
+
"gen_080",
|
| 284 |
+
"gen_082",
|
| 285 |
+
"gen_088",
|
| 286 |
+
"gen_091"
|
| 287 |
+
],
|
| 288 |
+
"seed": 42
|
| 289 |
+
}
|
unions.json
ADDED
|
The diff for this file is too large to render.
See raw diff
|
|
|