ner.rst

Named Entity Recognition

PyOBO has high-level wrappers to construct literal mapping objects defined by :mod:`ssslm.LiteralMapping`, which can be used to construct generic named entity recognition (NER) and named entity normalization (NEN) tooling (e.g., using ScispaCy or Gilda as a backend)

You can use :func:`pyobo.ground` as an integrated workflow:

import pyobo
import ssslm

matches: list[ssslm.Match] = pyobo.ground("taxrank", "species")

You can get the grounder directly first using :func:`pyobo.get_grounder`:

import pyobo
import ssslm

grounder: ssslm.Grounder = pyobo.get_grounder("taxrank")
matches: list[ssslm.Match] = grounder.get_matches("species")

You can get the ontology directly using :func:`pyobo.get_ontology` then construct a grounder with :meth:`pyobo.Obo.get_grounder`:

import pyobo
import ssslm

ontology: pyobo.Obo = pyobo.get_ontology("taxrank")
grounder: ssslm.Grounder = ontology.get_grounder()
matches: list[ssslm.Match] = grounder.get_matches("species")

You can load a custom ontology with :func:`pyobo.from_obo_path` then construct a grounder with :meth:`pyobo.Obo.get_grounder`:

import pyobo
import ssslm
from urllib.request import urlretrieve

url = "http://purl.obolibrary.org/obo/taxrank.obo"
path = "taxrank.obo"
urlretrieve(url, path)

ontology: pyobo.Obo = pyobo.from_obo_path(path, prefix="taxrank")
grounder: ssslm.Grounder = ontology.get_grounder()
matches: list[ssslm.Match] = grounder.get_matches("species")

Warning

When loading a custom ontology, it's required that the prefix is registered in the :mod:`bioregistry`, since PyOBO does additional standardization and normalization of prefixes, CURIEs, and URIs that are not part of the OBO specification.