Skip to main content

OntologyPorts

What it is​

  • Defines abstract “port” interfaces for ontology-based Named Entity Recognition (NER).
  • Establishes contracts for:
    • An NER adapter that maps text to an ontology (provided as Turtle).
    • A service that performs NER using an internal adapter/configured ontology.

Public API​

class IOntologyNERPort(ABC)​

Interface for an ontology-aware NER adapter.

  • named_entity_recognition(self, input: str, ontology_str: str) -> rdflib.Graph
    • Applies NER to input and maps recognized entities to ontology concepts defined in ontology_str (Turtle).
    • Returns an rdflib.Graph with mapped entities/relationships.

class IOntologyService(ABC)​

Interface for an ontology NER service.

  • named_entity_recognition(self, input: str) -> rdflib.Graph
    • Applies NER to input.
    • Returns an rdflib.Graph with mapped entities/relationships.

Configuration/Dependencies​

  • Dependencies
    • rdflib.Graph (return type for both ports)
    • langchain_core.language_models.BaseChatModel (typed private attribute in IOntologyNERPort)
  • Notes
    • The module defines private typed attributes (__chat_model, __ner_adaptor) but does not expose constructors or setters; concrete implementations are responsible for wiring these.

Usage​

Minimal example implementing both interfaces:

from rdflib import Graph
from naas_abi_core.services.ontology.OntologyPorts import IOntologyNERPort, IOntologyService

class DummyNER(IOntologyNERPort):
    def named_entity_recognition(self, input: str, ontology_str: str) -> Graph:
        return Graph()  # populate graph in a real implementation

class OntologyService(IOntologyService):
    def __init__(self, ner: IOntologyNERPort, ontology_str: str):
        self._ner = ner
        self._ontology_str = ontology_str

    def named_entity_recognition(self, input: str) -> Graph:
        return self._ner.named_entity_recognition(input, self._ontology_str)

svc = OntologyService(DummyNER(), ontology_str="")  # Turtle ontology string goes here
g = svc.named_entity_recognition("Some text")
print(len(g))

Caveats​

  • These are abstract interfaces only; no concrete NER logic is provided.
  • IOntologyNERPort.named_entity_recognition requires the ontology as a Turtle string (ontology_str).