Clinical Medical Text NLP Dataset

Dataset Overview

Dataset Name	Clinical medical text NLP Notified data set
Total Data Volume	328,000 (electronic medical records + diagnostic reports + medical advice + medical digests)
TextType	85K / M.C.S. / S.A. / S.A. / S.A. / S.A. / S.A. / S.A. / S.A. / S.A. / S. / / S.A. / S.A. / S. / S. / / / / /A. / / / /R. /
NER Entity Category	Diseases, symptoms, medicines, examinations, surgery, anatomy, laboratory indicators 7 major categories 46 subcategories
Relationship Description	Diseases - symptoms, drugs - diseases, detection - detection, surgery - autopsy
Standardization of terminology	ICD-10-CM / SNOMED CT / ATC / LOINC Map
Annotation Method	Clinical medicine professional label + NLP pre-marking aids + expert review of terminology
Format Output	BIO/BIOS Sequence Description / JSON / Brat Standoff / CNNLL

Detailed Description of Data Dimensions

Field Name	Data Type	Value Range / Description
doc_id	String(32)	De-sensitive Document ID
doc_type	Enum	ADMISSION \| PROGRESS \| DISCHARGE \| REPORT \| ORDER \| LITERATURE
department	Enum	INTERNAL \| SURGERY \| OBGYN \| PEDIATRICS \| ONCOLOGY \| CARDIOLOGY
text	Text	Clinical text after de-identification (patient name [NAME], date [DATE], institution [ORG])
entities	Array	[{start, end, type, subtype, text, icd10_code, snomed_code}]
relations	Array	[{head_entity_idx, tail_entity_idx, relation_type}]
bio_tags	Array	BIO/BIOS Sequence Tag List
text_length	Integer	Number of characters (incl. points)
entity_count	Integer	Number of entities
annotator_id	String(8)	Desensitized Marker ID
qc_score	Float	Quality assessment rating 0.0 - 1.0

De-identified JSON Data Sample

[ { "doc_id": "CLT-A0000821", "doc_type": "ADMISSION", "department": "CARDIOLOGY", "text": "[NAME]，[AGE][GENDER]，"、[DURATION]"。[DISEASE][DRUG]。：[VALUE]/，[VALUE]/[VALUE]mmHg。", "entities": [ {"start": 62, "end": 70, "type": "SYMPTOM", "subtype": "", "text": "、"}, {"start": 82, "end": 90, "type": "DISEASE", "subtype": "", "text": "[DISEASE]", "icd10_code": "I25.1"}, {"start": 92, "end": 98, "type": "DRUG", "subtype": "", "text": "[DRUG]", "atc_code": "C07AB"} ], "relations": [ {"head_entity_idx": 0, "tail_entity_idx": 1, "relation_type": "SYMPTOM_OF"} ], "text_length": 156, "entity_count": 3, "annotator_id": "ANN-0421", "qc_score": 0.96 }, { "doc_id": "CLT-R0012345", "doc_type": "REPORT", "department": "RADIOLOGY", "text": "：[BODY_PART]。Imaging Findings：[BODY_PART][SIZE]cm[DESCRIPTOR][FINDING]，[DEGREE]，[DESCRIPTOR]。：[DIAGNOSIS]。", "entities": [ {"start": 4, "end": 16, "type": "ANATOMY", "subtype": "", "text": "[BODY_PART]", "snomed_code": "123037004"}, {"start": 20, "end": 32, "type": "ANATOMY", "subtype": "", "text": "[BODY_PART]"}, {"start": 33, "end": 42, "type": "LAB", "subtype": "", "text": "[SIZE]cm"}, {"start": 43, "end": 55, "type": "FINDING", "subtype": "ImagingDiscovery", "text": "[DESCRIPTOR][FINDING]"} ], "relations": [ {"head_entity_idx": 1, "tail_entity_idx": 3, "relation_type": "HAS_FINDING"} ], "text_length": 89, "entity_count": 4, "annotator_id": "ANN-0715", "qc_score": 0.94 }, { "doc_id": "CLT-D0003672", "doc_type": "DISCHARGE", "department": "ONCOLOGY", "text": "：[DISEASE]（[STAGE]）。[PROCEDURE]，。：1.[DRUG] [DOSAGE]；2.[LAB]；3.。", "entities": [ {"start": 5, "end": 15, "type": "DISEASE", "subtype": "", "text": "[DISEASE]", "icd10_code": "C34.9"}, {"start": 24, "end": 32, "type": "PROCEDURE", "subtype": "", "text": "[PROCEDURE]"}, {"start": 47, "end": 53, "type": "DRUG", "subtype": "", "text": "[DRUG]", "atc_code": "L01XE"}, {"start": 64, "end": 70, "type": "LAB", "subtype": "", "text": "[LAB]", "loinc_code": "24331-1"} ], "relations": [ {"head_entity_idx": 0, "tail_entity_idx": 1, "relation_type": "TREATED_BY"}, {"head_entity_idx": 0, "tail_entity_idx": 3, "relation_type": "MONITORED_BY"} ], "text_length": 112, "entity_count": 4, "annotator_id": "ANN-0388", "qc_score": 0.97 }, { "doc_id": "CLT-O0009105", "doc_type": "ORDER", "department": "INTERNAL", "text": "[DRUG] [DOSAGE] [ROUTE] [FREQUENCY] [DURATION]", "entities": [ {"start": 0, "end": 6, "type": "DRUG", "subtype": "", "text": "[DRUG]", "atc_code": "J01MA"}, {"start": 7, "end": 15, "type": "DOSAGE", "subtype": "", "text": "[DOSAGE]"}, {"start": 16, "end": 22, "type": "ROUTE", "subtype": "", "text": "[ROUTE]"} ], "relations": [], "text_length": 42, "entity_count": 3, "annotator_id": "ANN-0912", "qc_score": 0.99 }, { "doc_id": "CLT-L0001543", "doc_type": "LITERATURE", "department": "ONCOLOGY", "text": "Research[DISEASE][DRUG][DRUG]。[NUMBER]，ORR[VALUE]%，PFS[VALUE]。", "entities": [ {"start": 6, "end": 16, "type": "DISEASE", "subtype": "", "text": "[DISEASE]"}, {"start": 18, "end": 24, "type": "DRUG", "subtype": "", "text": "[DRUG]"}, {"start": 26, "end": 32, "type": "DRUG", "subtype": "", "text": "[DRUG]"}, {"start": 40, "end": 48, "type": "LAB", "subtype": "Clinical", "text": "ORR"}, {"start": 55, "end": 63, "type": "LAB", "subtype": "Clinical", "text": "PFS"} ], "relations": [ {"head_entity_idx": 0, "tail_entity_idx": 1, "relation_type": "TREATED_BY"}, {"head_entity_idx": 0, "tail_entity_idx": 3, "relation_type": "MEASURED_BY"} ], "text_length": 97, "entity_count": 5, "annotator_id": "ANN-0523", "qc_score": 0.95 } ]

Annotation Specifications and Quality Control

Annotation Standards

NER label following i2b2 2010/2012 Extension of the Clinical Named Entity Identification Standard, adding 46 fine particle categories
Relationship extraction is in the three-tier group format of the "head entity relationship tailing entity" and relationship types cover diagnostic-discovery, treatment-disease, etc.
Standardized dual map of terms: ICD-10-CM diagnostic code + SNOMED CT clinical term, ATC code for pharmaceuticals
Text de-identification: PHI (Proteced Health Information) was replaced by a type placeholder, zero information leak

Quality Metrics

96.3%

NER F1

91.7%

Relationship extraction F1

98.5%

The term map accuracy rate

0.89

Mark Consistency Rate Kappa

AI Training Use Cases

Clinical naming entity identification (NER)

Applying model: BERT-BiLSTM-CRF / BioBERT / PubMedBERT / Clinical BERT / GatorTron. Trained in the name entity identification model for Chinese clinical text, F1 can reach 90%+.

Medical relations extraction

Applying model: BioRelEx / PURE / PL-Marker / REBEL. Draws the disease-symptoms, pharmaceuticals-disease relationships on the basis of physical identification, and constructs a medical knowledge map.

Standardization of clinical terminology

Applying model: SapBERT / CODER / BioSyn. Map oral/non-standard clinical terms to the ICD-10/SNOMED CT standard terminology system.

Electronic medical history structured

Applying model: LLM + RAG (GPT-4o/Claude/Qwen2.5 + Langchain). End-to-end conversion of unstructured medical records to structured FHIR/CDA format.

Get This Dataset

The clinical NLP data sets are flexiblely authorized by the type of section/marking, supporting customisation of labelling requirements.

137-5502-0164 Online Consultation

Related Datasets

Commercial Application Value

Informationalization of Smart Hospitals

Smart patient history control embedded in the HIS/EMR system, CDSS clinical decision support module to improve patient record regulation and quality of treatment.

Medical intelligence clearance

Based on diagnostic-pharmacological-check relationships, training in the mental-censorship model, automatic identification of unreasonable use and over-checking.

The Medicine Real World Study

Structured drug information and clinical results from a large electronic medical record to support the RWE study on real world evidence after drugs are listed.