Datasetpublic

Sidharth1743/indicphi-clinical-ner-500

IndicPHI Synthetic Clinical NER (GLiNER) Synthetic multilingual Indian clinical documents for NER / GLiNER training, with PHI/PII surrogate entities. Produced by IndicPHI (Track C SDG). Provenance Source run: 20260724T100211 Artifact: data/generated/runs/20260724T100211/s9_gliner_s6_export/gliner_docs.jsonl Rows: 479 Languages: 23 Language counts code n as 22 bn 20 brx 19 doi 21 en 22 gu 21 hi 22 kn 22 kok 21 ks 21… See the full description on the dataset page: https://huggingface.co/datasets/Sidharth1743/indicphi-clinical-ner-500.

sourceHugging Faceapache-2.0updated 2mo agoView on Hugging Face

0likes20downloads

Dataset Card

IndicPHI Synthetic Clinical NER (GLiNER)

Synthetic multilingual Indian clinical documents for NER / GLiNER training, with PHI/PII surrogate entities. Produced by IndicPHI (Track C SDG).

Provenance

—Source run: 20260724T100211
—Artifact: data/generated/runs/20260724T100211/s9_gliner_s6_export/gliner_docs.jsonl
—Rows: 479
—Languages: 23

Language counts

code	n
`as`	22
`bn`	20
`brx`	19
`doi`	21
`en`	22
`gu`	21
`hi`	22
`kn`	22
`kok`	21
`ks`	21
`mai`	22
`ml`	21
`mni`	21
`mr`	21
`ne`	21
`or`	22
`pa`	21
`sa`	21
`sat`	17
`sd`	16
`ta`	22
`te`	21
`ur`	22

NER schema (GLiNER-compatible)

Each row is ready for token-classification / GLiNER training:

json

{
  "tokenized_text": ["…", "…"],
  "ner": [{"start": 0, "end": 1, "label": "PATIENT_NAME"}, …],
  "ner_gliner_json": "[[0, 1, \"PATIENT_NAME\"], …]",
  "text": "detagged surface text",
  "document_language_code": "hi"
}

ner uses inclusive token indices as Arrow-safe structs. For classic GLiNER triples, json.loads(row["ner_gliner_json"]).

Intended use

Train or evaluate Indic clinical NER / GLiNER models. Synthetic only — not real PHI.