Add dataset card with documentation and multi-config definition

README.md

@@ -0,0 +1,203 @@
+---
+license: cc0-1.0
+language:
+  - en
+tags:
+  - metadata
+  - scholarly-communication
+  - datacite
+  - doi
+  - enrichment
+  - comet
+pretty_name: COMET DataCite Enrichment Layer (Ingest Format)
+size_categories:
+  - 10M<n<100M
+configs:
+  - config_name: author_affiliations
+    data_files:
+      - split: train
+        path: author_affiliations_enrichments_datacite_format.jsonl
+    default: true
+  - config_name: preprint_matching
+    data_files:
+      - split: train
+        path: preprint_matching_enrichments_datacite_format.jsonl
+  - config_name: resource_type
+    data_files:
+      - split: train
+        path: resource_type_enrichments_datacite_format.jsonl
+---
+
+# COMET DataCite Enrichment Layer (Ingest Format)
+
+This dataset contains metadata enrichment records in [DataCite](https://datacite.org/) ingest format, produced by the [COMET](https://cometadata.org/) (Collaborative Metadata) initiative. Each record represents a single metadata enrichment to be applied to an existing DataCite record, identified by DOI.
+
+The enrichments were generated against the **February 2026** DataCite data files using the [datacite-enrichment](https://github.com/adambuttrick/datacite-enrichment) tool.
+
+## Dataset Configs
+
+This dataset contains three configs (subsets), one for each enrichment type:
+
+| Config | Records | Size | Description |
+|--------|---------|------|-------------|
+| `author_affiliations` | 8,963,119 | 8.8 GB | Institutional affiliation enrichments for creator records |
+| `preprint_matching` | 812,958 | 526 MB | Links between arXiv preprints and their published versions |
+| `resource_type` | 3,338,250 | 2.1 GB | Reclassifications of generic resource types to specific types |
+
+### Loading a specific config
+
+```python
+from datasets import load_dataset
+
+# Load author affiliations enrichments
+ds = load_dataset("cometadata/comet-datacite-enrichment-layer-ingest-format", "author_affiliations")
+
+# Load preprint matching enrichments
+ds = load_dataset("cometadata/comet-datacite-enrichment-layer-ingest-format", "preprint_matching")
+
+# Load resource type enrichments
+ds = load_dataset("cometadata/comet-datacite-enrichment-layer-ingest-format", "resource_type")
+```
+
+## Record Schema
+
+Each record is a JSON object with the following fields:
+
+| Field | Type | Description |
+|-------|------|-------------|
+| `doi` | string | The DOI of the DataCite record to be enriched |
+| `contributors` | array | Provenance: who produced this enrichment (COMET as Producer) |
+| `resources` | array | Provenance: related project DOI and source dataset reference |
+| `action` | string | The enrichment action: `insert`, `updateChild`, or `update` |
+| `field` | string | The DataCite metadata field being enriched |
+| `originalValue` | object | The original value being modified (present for `updateChild` and `update` actions) |
+| `enrichedValue` | object | The new or updated value to apply |
+
+## Config Details
+
+### author_affiliations
+
+Adds institutional affiliations to creator records on arXiv preprints registered in DataCite.
+
+- **Action**: `updateChild` — finds and updates a specific creator within the `creators` array
+- **Field**: `creators`
+- **Source data**: [cometadata/arxiv-author-affiliations-matched-ror-ids](https://huggingface.co/datasets/cometadata/arxiv-author-affiliations-matched-ror-ids)
+- **Project DOI**: [10.82461/160e-8q92](https://doi.org/10.82461/160e-8q92)
+
+Example record:
+```json
+{
+  "doi": "10.48550/arxiv.1109.3791",
+  "contributors": [
+    {"name": "COMET", "contributorType": "Producer", "nameType": "Organizational"}
+  ],
+  "resources": [
+    {"relatedIdentifier": "http://doi.org/10.82461/160e-8q92", "relatedIdentifierType": "DOI", "relationType": "IsDocumentedBy", "resourceTypeGeneral": "Project"},
+    {"relatedIdentifier": "https://huggingface.co/datasets/cometadata/arxiv-author-affiliations-matched-ror-ids", "relatedIdentifierType": "URL", "relationType": "IsDerivedFrom", "resourceTypeGeneral": "Dataset"}
+  ],
+  "action": "updateChild",
+  "field": "creators",
+  "originalValue": {
+    "nameType": "Personal",
+    "givenName": "Fangfei",
+    "familyName": "Zhou",
+    "name": "Zhou, Fangfei"
+  },
+  "enrichedValue": {
+    "nameType": "Personal",
+    "givenName": "Fangfei",
+    "familyName": "Zhou",
+    "name": "Zhou, Fangfei",
+    "affiliation": [
+      {"name": "Northeastern University"}
+    ]
+  }
+}
+```
+
+### preprint_matching
+
+Links arXiv preprints to their published versions by inserting `relatedIdentifiers` entries with the `IsVersionOf` relation type.
+
+- **Action**: `insert` — appends a new entry to the `relatedIdentifiers` array
+- **Field**: `relatedIdentifiers`
+- **Source data**: [cometadata/arxiv-preprint-matching-results](https://huggingface.co/datasets/cometadata/arxiv-preprint-matching-results)
+- **Project DOI**: [10.82461/m8a8-m211](https://doi.org/10.82461/m8a8-m211)
+
+Example record:
+```json
+{
+  "doi": "10.48550/arxiv.2302.11570",
+  "contributors": [
+    {"name": "COMET", "contributorType": "Producer", "nameType": "Organizational"}
+  ],
+  "resources": [
+    {"relatedIdentifier": "10.82461/m8a8-m211", "relatedIdentifierType": "DOI", "relationType": "IsDocumentedBy", "resourceTypeGeneral": "Project"},
+    {"relatedIdentifier": "https://huggingface.co/datasets/cometadata/arxiv-preprint-matching-results", "relatedIdentifierType": "DOI", "relationType": "IsDerivedFrom", "resourceTypeGeneral": "Dataset"}
+  ],
+  "action": "insert",
+  "field": "relatedIdentifiers",
+  "enrichedValue": {
+    "relatedIdentifier": "10.58530/2024/0038",
+    "relatedIdentifierType": "DOI",
+    "relationType": "IsVersionOf"
+  }
+}
+```
+
+### resource_type
+
+Reclassifies DataCite records that have a generic `resourceTypeGeneral` of "Text" to more specific types such as `JournalArticle`, `Preprint`, etc., based on procedural analysis.
+
+- **Action**: `update` — replaces the entire `types` field
+- **Field**: `types`
+- **Source data**: [cometadata/datacite-procedural-resource-type-general-reclassifications](https://huggingface.co/datasets/cometadata/datacite-procedural-resource-type-general-reclassifications)
+- **Project DOI**: [10.82461/bpzr-jd55](https://doi.org/10.82461/bpzr-jd55)
+
+Example record:
+```json
+{
+  "doi": "10.5281/zenodo.15414",
+  "contributors": [
+    {"name": "COMET", "contributorType": "Producer", "nameType": "Organizational"}
+  ],
+  "resources": [
+    {"relatedIdentifier": "10.82461/bpzr-jd55", "relatedIdentifierType": "DOI", "relationType": "IsDocumentedBy", "resourceTypeGeneral": "Project"},
+    {"relatedIdentifier": "https://huggingface.co/datasets/cometadata/datacite-procedural-resource-type-general-reclassifications", "relatedIdentifierType": "URL", "relationType": "IsDerivedFrom", "resourceTypeGeneral": "Dataset"}
+  ],
+  "action": "update",
+  "field": "types",
+  "originalValue": {"resourceTypeGeneral": "Text"},
+  "enrichedValue": {"resourceTypeGeneral": "JournalArticle"}
+}
+```
+
+## How These Files Were Generated
+
+These enrichment records were produced using the [datacite-enrichment](https://github.com/adambuttrick/datacite-enrichment) tool, which:
+
+1. Takes source enrichment data (from the datasets linked above) and a YAML configuration file defining field mappings, filters, and provenance metadata
+2. Converts it to a standardized enrichment format using domain-specific transformers (one per enrichment type)
+3. Produces records that carry full provenance tracking: contributor info (COMET as Producer), references to the source dataset and project DOI, and original/enriched values
+
+The conversion was run against the **February 2026** DataCite PID Graph data files. Each record in this dataset is ready to be applied to its corresponding DataCite record via the tool's merge functionality, or ingested directly by DataCite.
+
+### Generation commands
+
+```bash
+# Preprint matching
+datacite-enrich convert preprint_matches.json configs/preprint_matching.yaml \
+  -r datacite_records.jsonl.gz -o preprint_matching_enrichments_datacite_format.jsonl
+
+# Author affiliations
+datacite-enrich convert author_affiliations.jsonl configs/author_affiliations.yaml \
+  -r datacite_records.jsonl.gz -o author_affiliations_enrichments_datacite_format.jsonl
+
+# Resource type
+datacite-enrich convert resource_type_reclassifications.json configs/resource_type.yaml \
+  -r datacite_records.jsonl.gz -o resource_type_enrichments_datacite_format.jsonl
+```
+
+## License
+
+This dataset is released under [CC0 1.0](https://creativecommons.org/publicdomain/zero/1.0/).