Datasets:
v1.0.1: per-split Parquet files mapped to train/validation/test, leaderboard rule, fixture_pairs config fixed
Browse files- README.md +67 -14
- data/pairs-dev.parquet +3 -0
- data/pairs-test.parquet +3 -0
- data/pairs-train.parquet +3 -0
- data/records-dev.parquet +3 -0
- data/records-test.parquet +3 -0
- data/records-train.parquet +3 -0
- fixture/pairs-dev.parquet +3 -0
- fixture/pairs-test.parquet +3 -0
- fixture/pairs-train.parquet +3 -0
- fixture/records-dev.parquet +3 -0
- fixture/records-test.parquet +3 -0
- fixture/records-train.parquet +3 -0
README.md
CHANGED
|
@@ -19,15 +19,27 @@ configs:
|
|
| 19 |
default: true
|
| 20 |
data_files:
|
| 21 |
- split: train
|
| 22 |
-
path: data/pairs.parquet
|
|
|
|
|
|
|
|
|
|
|
|
|
| 23 |
- config_name: records
|
| 24 |
data_files:
|
| 25 |
- split: train
|
| 26 |
-
path: data/records.parquet
|
|
|
|
|
|
|
|
|
|
|
|
|
| 27 |
- config_name: fixture_pairs
|
| 28 |
data_files:
|
| 29 |
- split: train
|
| 30 |
-
path: fixture/pairs.
|
|
|
|
|
|
|
|
|
|
|
|
|
| 31 |
---
|
| 32 |
|
| 33 |
# USORG-Bench
|
|
@@ -38,13 +50,15 @@ registration number, USAC billed-entity or service-provider number, or state bus
|
|
| 38 |
number, and do not match when they carry different values of the same identifier.
|
| 39 |
|
| 40 |
* 62,497 labelled pairs (26,149 matches, 36,348 non-matches) over 96,642 records and 71,783
|
| 41 |
-
|
| 42 |
subawards), SEC EDGAR, CMS hospital data, FCC ULS and LMS, USAC E-Rate, and the Colorado,
|
| 43 |
Connecticut, New York and Oregon business registries.
|
| 44 |
* Organizations only: sole proprietorships and person names are removed.
|
| 45 |
* Identifier values are masked (`entity_id` is a keyed HMAC); names and business addresses are
|
| 46 |
as the sources print them.
|
| 47 |
-
*
|
|
|
|
|
|
|
| 48 |
* `difficulty` strata (renames and reworded names are hard matches; near-identical names of
|
| 49 |
different organizations are hard non-matches) and a `label_confidence` column.
|
| 50 |
* A 300-pair audit (`audit/audit_300.csv`).
|
|
@@ -52,13 +66,35 @@ number, and do not match when they carry different values of the same identifier
|
|
| 52 |
Code, loader, metrics, baselines and the build scripts: https://github.com/Mattbusel/usorg-bench
|
| 53 |
(`pip install usorg-bench` once published; until then `pip install git+https://github.com/Mattbusel/usorg-bench`).
|
| 54 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 55 |
## Files
|
| 56 |
|
| 57 |
| path | contents |
|
| 58 |
|---|---|
|
| 59 |
| `data/records.{csv,parquet}` | one row per source record |
|
| 60 |
| `data/pairs.{csv,parquet}` | labelled pairs with split, difficulty, confidence, categories |
|
| 61 |
-
| `
|
|
|
|
| 62 |
| `audit/audit_300.csv` | 300 audited test pairs with verdicts |
|
| 63 |
| `baselines/` | baseline scores (`pair_id,score`) and `results.json` |
|
| 64 |
| `exports/name_bench/` | 300 rows in nomenklatura's `contrib/name_bench/cases.csv` format |
|
|
@@ -71,12 +107,21 @@ Datasheet: [docs/datasheet.md](https://github.com/Mattbusel/usorg-bench/blob/mai
|
|
| 71 |
|
| 72 |
## Quick start
|
| 73 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 74 |
```python
|
| 75 |
import pandas as pd
|
| 76 |
-
|
| 77 |
-
|
| 78 |
-
|
| 79 |
-
|
|
|
|
| 80 |
```
|
| 81 |
|
| 82 |
or with the package:
|
|
@@ -84,13 +129,12 @@ or with the package:
|
|
| 84 |
```python
|
| 85 |
import usorg_bench
|
| 86 |
data = usorg_bench.load("full") # downloads and checks this revision
|
| 87 |
-
|
| 88 |
-
for pair, a, b in test.iter_named_pairs():
|
| 89 |
...
|
| 90 |
```
|
| 91 |
|
| 92 |
Score each pair from 0 to 1, pick the threshold on `dev`, report F1 and average precision on
|
| 93 |
-
`test` (`usorg-bench eval scores.csv` does this).
|
| 94 |
|
| 95 |
## Baselines (test split, threshold chosen on dev)
|
| 96 |
|
|
@@ -100,11 +144,12 @@ Score each pair from 0 to 1, pick the threshold on `dev`, report F1 and average
|
|
| 100 |
| logicv2_full | name + address | 0.624 | 0.561 | 0.703 | 0.628 | 0.6989 |
|
| 101 |
| logicv2 | name | 0.588 | 0.416 | 1.000 | 0.510 | 0.0000 |
|
| 102 |
| rapidfuzz | name | 0.588 | 0.416 | 1.000 | 0.498 | 0.0000 |
|
|
|
|
| 103 |
|
| 104 |
Name similarity alone does not separate the classes by design: hard non-matches are
|
| 105 |
organizations with the same or nearly the same name (two towns called Springfield), and hard
|
| 106 |
matches are renamed or reworded organizations. A method that reads the address fields can tell
|
| 107 |
-
many of the same-name pairs apart.
|
| 108 |
|
| 109 |
## Audit
|
| 110 |
|
|
@@ -113,6 +158,14 @@ many of the same-name pairs apart.
|
|
| 113 |
The release filters written after that review remove the 5 error patterns (503 pairs in all); of
|
| 114 |
the 293 audited pairs left in the release, 285 agree and 8 are unclear. Details in the datasheet.
|
| 115 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 116 |
## License and citation
|
| 117 |
|
| 118 |
CC BY 4.0. The underlying federal records are public domain; state registry records are public
|
|
|
|
| 19 |
default: true
|
| 20 |
data_files:
|
| 21 |
- split: train
|
| 22 |
+
path: data/pairs-train.parquet
|
| 23 |
+
- split: validation
|
| 24 |
+
path: data/pairs-dev.parquet
|
| 25 |
+
- split: test
|
| 26 |
+
path: data/pairs-test.parquet
|
| 27 |
- config_name: records
|
| 28 |
data_files:
|
| 29 |
- split: train
|
| 30 |
+
path: data/records-train.parquet
|
| 31 |
+
- split: validation
|
| 32 |
+
path: data/records-dev.parquet
|
| 33 |
+
- split: test
|
| 34 |
+
path: data/records-test.parquet
|
| 35 |
- config_name: fixture_pairs
|
| 36 |
data_files:
|
| 37 |
- split: train
|
| 38 |
+
path: fixture/pairs-train.parquet
|
| 39 |
+
- split: validation
|
| 40 |
+
path: fixture/pairs-dev.parquet
|
| 41 |
+
- split: test
|
| 42 |
+
path: fixture/pairs-test.parquet
|
| 43 |
---
|
| 44 |
|
| 45 |
# USORG-Bench
|
|
|
|
| 50 |
number, and do not match when they carry different values of the same identifier.
|
| 51 |
|
| 52 |
* 62,497 labelled pairs (26,149 matches, 36,348 non-matches) over 96,642 records and 71,783
|
| 53 |
+
distinct identifiers (one organization can hold several), from 15 public sources: USAspending (FPDS contracts, recipient profiles, DoD
|
| 54 |
subawards), SEC EDGAR, CMS hospital data, FCC ULS and LMS, USAC E-Rate, and the Colorado,
|
| 55 |
Connecticut, New York and Oregon business registries.
|
| 56 |
* Organizations only: sole proprietorships and person names are removed.
|
| 57 |
* Identifier values are masked (`entity_id` is a keyed HMAC); names and business addresses are
|
| 58 |
as the sources print them.
|
| 59 |
+
* `train` / `dev` / `test` splits disjoint by identifier (31,199 / 6,123 / 25,175 pairs). In
|
| 60 |
+
`datasets.load_dataset` they are `train`, `validation` and `test`; records of a split are the
|
| 61 |
+
ones its pairs reference.
|
| 62 |
* `difficulty` strata (renames and reworded names are hard matches; near-identical names of
|
| 63 |
different organizations are hard non-matches) and a `label_confidence` column.
|
| 64 |
* A 300-pair audit (`audit/audit_300.csv`).
|
|
|
|
| 66 |
Code, loader, metrics, baselines and the build scripts: https://github.com/Mattbusel/usorg-bench
|
| 67 |
(`pip install usorg-bench` once published; until then `pip install git+https://github.com/Mattbusel/usorg-bench`).
|
| 68 |
|
| 69 |
+
## Leaderboard rule: allowed inputs
|
| 70 |
+
|
| 71 |
+
A reported result may read only `name`, `address`, `city`, `state` and `zip5` (and the ids as
|
| 72 |
+
opaque keys). `source`, `id_type`, `entity_id`, `name_kind` and the pair columns (`difficulty`,
|
| 73 |
+
`categories`, `name_sim`, `label_confidence`) follow from how the pairs were built and predict the
|
| 74 |
+
label without reading a name. Measured on the test split with a lookup table fit on `train` and
|
| 75 |
+
the threshold picked on `dev`:
|
| 76 |
+
|
| 77 |
+
| key looked up | test F1 | test AP |
|
| 78 |
+
|---|---|---|
|
| 79 |
+
| `id_type` + the two records' `source` | 0.927 | 0.908 |
|
| 80 |
+
| the same + state equal + names identical | 0.973 | 0.980 |
|
| 81 |
+
| `difficulty` + `categories` | 0.923 | 0.976 |
|
| 82 |
+
| state equal (an allowed field) | 0.800 | 0.680 |
|
| 83 |
+
|
| 84 |
+
The last line is a property of the sampling: the build drops same-state near-name non-matches,
|
| 85 |
+
so "same state" alone is worth F1 0.800. Read record-level results against that, not against the
|
| 86 |
+
name-only all-match F1 of 0.588. Also by construction: every pair whose names are identical after
|
| 87 |
+
case and punctuation folding is a non-match (2,113 pairs, all but one in different states).
|
| 88 |
+
Details: [docs/metrics.md](https://github.com/Mattbusel/usorg-bench/blob/main/docs/metrics.md#allowed-inputs).
|
| 89 |
+
|
| 90 |
## Files
|
| 91 |
|
| 92 |
| path | contents |
|
| 93 |
|---|---|
|
| 94 |
| `data/records.{csv,parquet}` | one row per source record |
|
| 95 |
| `data/pairs.{csv,parquet}` | labelled pairs with split, difficulty, confidence, categories |
|
| 96 |
+
| `data/pairs-{train,dev,test}.parquet`, `data/records-{train,dev,test}.parquet` | the same rows, one file per split (since `v1.0.1`) |
|
| 97 |
+
| `fixture/` | the 5,000-pair subset bundled with the Python package (CSV, and Parquet per split) |
|
| 98 |
| `audit/audit_300.csv` | 300 audited test pairs with verdicts |
|
| 99 |
| `baselines/` | baseline scores (`pair_id,score`) and `results.json` |
|
| 100 |
| `exports/name_bench/` | 300 rows in nomenklatura's `contrib/name_bench/cases.csv` format |
|
|
|
|
| 107 |
|
| 108 |
## Quick start
|
| 109 |
|
| 110 |
+
```python
|
| 111 |
+
from datasets import load_dataset
|
| 112 |
+
pairs = load_dataset("tensorust/usorg-bench", "pairs", revision="v1.0.1") # train / validation / test
|
| 113 |
+
records = load_dataset("tensorust/usorg-bench", "records", revision="v1.0.1")
|
| 114 |
+
```
|
| 115 |
+
|
| 116 |
+
or with pandas:
|
| 117 |
+
|
| 118 |
```python
|
| 119 |
import pandas as pd
|
| 120 |
+
base = "hf://datasets/tensorust/usorg-bench@v1.0.1/data"
|
| 121 |
+
test = pd.read_parquet(f"{base}/pairs-test.parquet")
|
| 122 |
+
records = pd.read_parquet(f"{base}/records.parquet").set_index("record_id")
|
| 123 |
+
allowed = ["name", "address", "city", "state", "zip5"]
|
| 124 |
+
left, right = records.loc[test.left_id, allowed], records.loc[test.right_id, allowed]
|
| 125 |
```
|
| 126 |
|
| 127 |
or with the package:
|
|
|
|
| 129 |
```python
|
| 130 |
import usorg_bench
|
| 131 |
data = usorg_bench.load("full") # downloads and checks this revision
|
| 132 |
+
for pair_id, a, b in data.select("test").iter_feature_pairs(): # allowed fields only
|
|
|
|
| 133 |
...
|
| 134 |
```
|
| 135 |
|
| 136 |
Score each pair from 0 to 1, pick the threshold on `dev`, report F1 and average precision on
|
| 137 |
+
`test` (`usorg-bench eval scores.csv` does this; include the dev scores in the file).
|
| 138 |
|
| 139 |
## Baselines (test split, threshold chosen on dev)
|
| 140 |
|
|
|
|
| 144 |
| logicv2_full | name + address | 0.624 | 0.561 | 0.703 | 0.628 | 0.6989 |
|
| 145 |
| logicv2 | name | 0.588 | 0.416 | 1.000 | 0.510 | 0.0000 |
|
| 146 |
| rapidfuzz | name | 0.588 | 0.416 | 1.000 | 0.498 | 0.0000 |
|
| 147 |
+
| reference: state equal | state | 0.800 | 0.696 | 0.941 | 0.680 | 0.6966 |
|
| 148 |
|
| 149 |
Name similarity alone does not separate the classes by design: hard non-matches are
|
| 150 |
organizations with the same or nearly the same name (two towns called Springfield), and hard
|
| 151 |
matches are renamed or reworded organizations. A method that reads the address fields can tell
|
| 152 |
+
many of the same-name pairs apart, but part of that is the same-state sampling effect above.
|
| 153 |
|
| 154 |
## Audit
|
| 155 |
|
|
|
|
| 158 |
The release filters written after that review remove the 5 error patterns (503 pairs in all); of
|
| 159 |
the 293 audited pairs left in the release, 285 agree and 8 are unclear. Details in the datasheet.
|
| 160 |
|
| 161 |
+
## Revisions
|
| 162 |
+
|
| 163 |
+
* `v1.0` (2026-11-17): first release.
|
| 164 |
+
* `v1.0.1` (2026-11-17): the same rows; adds one Parquet file per split and maps them to
|
| 165 |
+
`train` / `validation` / `test` (in `v1.0` every pair sat in one `train` split of the viewer),
|
| 166 |
+
fixes the `fixture_pairs` config, and adds the leaderboard rule above. `usorg-bench` 0.1.1
|
| 167 |
+
loads this revision.
|
| 168 |
+
|
| 169 |
## License and citation
|
| 170 |
|
| 171 |
CC BY 4.0. The underlying federal records are public domain; state registry records are public
|
data/pairs-dev.parquet
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:4bff12e848ce503e63c99be8d1df2d6e01e1f1c704749e7702c686b39a4520ad
|
| 3 |
+
size 267294
|
data/pairs-test.parquet
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:7f76a6005d6d1d1d7e332c390178c840e9752ad49d08f416c43829688e008e6b
|
| 3 |
+
size 1072851
|
data/pairs-train.parquet
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:e5078ddc36fa880bc2110d7ba186d17b46177e3332fdcd761a58ebaa45ae4ee2
|
| 3 |
+
size 1326782
|
data/records-dev.parquet
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:7c1b1912d1dba6c8359d1029a63d12e544a11864752fce59d9e795f3203e6f1d
|
| 3 |
+
size 664232
|
data/records-test.parquet
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:a381103136fb8f2e8017f1c0945f75b2fb826e4caba25e224efb3ba7f5bf70a9
|
| 3 |
+
size 2638606
|
data/records-train.parquet
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:b771fa9251ada74ec9d6eaf97df953b02027dc079689239862c5889b7540c07d
|
| 3 |
+
size 3261005
|
fixture/pairs-dev.parquet
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:036e6595703405a69bc83bd108e3db9cd2c4f96e463e935aeb22dedf650ce747
|
| 3 |
+
size 29614
|
fixture/pairs-test.parquet
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:5c12ca7e13d1f6e36c7463a123e87a6e7ca9991f1de1d8dd4e3493e7053c7153
|
| 3 |
+
size 99890
|
fixture/pairs-train.parquet
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:e7cad39dff83500c385ebfd9d8303890f7f019f9d32623b746efa82f6bb1557b
|
| 3 |
+
size 122808
|
fixture/records-dev.parquet
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:06425c445618fef7123a6760ec8278f15724417c44b8234890d871fe24f75a0d
|
| 3 |
+
size 77320
|
fixture/records-test.parquet
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:b1a646b98cde75797dd3a59738f87349fc4556519ec74a1aaea52e9ed84d74de
|
| 3 |
+
size 287641
|
fixture/records-train.parquet
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:d5e2f70c76653ee01872af7fdcf76894c9bb8dc7baa67659697780c272f90915
|
| 3 |
+
size 353562
|