Saltar al contenido principal

NER Extractor

Category: AI & ML
Node id: ner-extractor

Standards: HIPAA

Performs Named Entity Recognition on clinical text using a dedicated NER service or LLM fallback

What it does​

  • Clinical Named Entity Recognition (NER)
  • Dedicated NER service integration with LLM fallback
  • Configurable entity type filtering
  • Entity position and confidence scoring
  • Support for clinical entity types (CONDITION, MEDICATION, PROCEDURE, ANATOMY, etc.)

Use cases​

  • Extracting diagnoses and medications from clinical notes
  • Identifying procedures and anatomical references in discharge summaries
  • Building structured data from unstructured clinical text
  • Pre-processing text for FHIR resource generation

Runtime​

  • Performance tier: variable

Input ports​

NameTypeRequiredDescription
texttextYesClinical text to extract entities from

Output ports​

NameTypeRequiredDescription
entitiesjsonNoExtracted named entities with type, position, and confidence
sourcetextNoSource used for extraction (ner_service or llm)
http_statusanyNoHTTP status code from the service

Settings​

Basic​

SettingNameTypeRequiredDefaultAllowed valuesDescription
NER Endpointner_endpointstringNoBase URL of the dedicated NER service
LLM Endpointllm_endpointstringNoBase URL of the LLM API (fallback if no NER endpoint)
ModelmodelstringNoNER model name

Auth​

SettingNameTypeRequiredDefaultAllowed valuesDescription
API Keyapi_keysecretNoAPI key for the service

Advanced​

SettingNameTypeRequiredDefaultAllowed valuesDescription
Entity Typesentity_typesstringNoallTypes of entities to extract (comma-separated or 'all')

Errors​

CodeDescription
TEXT_REQUIREDNo text input was provided for entity extraction
ENDPOINT_REQUIREDNeither ner_endpoint nor llm_endpoint is configured
NER_SERVICE_FAILEDDedicated NER service request failed
HTTP_ERRORNER or LLM service returned an HTTP error status

Runtime notes​

  • Output shape differs by source: NER service returns the raw service response with 'source' injected; LLM fallback returns {entities, source, http_status}
  • LLM fallback parses the model response as JSON; if parsing fails, entities is a raw string
  • Entity types include CONDITION, MEDICATION, PROCEDURE, ANATOMY, LAB_TEST, LAB_VALUE, VITAL_SIGN, ALLERGY, SYMPTOM, DEVICE
  • Requires either ner_endpoint or llm_endpoint; the node prefers the dedicated NER service when both are set

See Build any workflow for how to put nodes together and Ports, types and data flow for how port types connect.

¿Algo no está claro? Envíe sus comentarios sobre la documentación por correo electrónico..