Basic matcher terminology fails

[fliot]

Basic matcher terminology example fails,
Bug or coding error ?

import edsnlp
import edsnlp.pipes as eds

nlp = edsnlp.blank("eds")

nlp.add_pipe(eds.normalizer())
nlp.add_pipe("sentencizer")

terms = {
    "diabete": ["diabète", "diabete", "diabetes"],
    "hta": ["hypertension artérielle", "HTA"]
}

nlp.add_pipe(
    eds.matcher(
        terms=terms,
        attr="NORM",
        term_matcher="exact"
    ),
    name="terminology"
)

nlp.add_pipe(eds.negation())
nlp.add_pipe(eds.family())
nlp.add_pipe(eds.hypothesis())

print("Pipeline components:", nlp.pipe_names)

doc = nlp("Pas de diabète. Le père a une hypertension artérielle.")

print("Available spans:", doc.spans.keys())

for ent in doc.spans["terminology"]:
    print(ent.text, ent.label_, ent._.negation_, ent._.family_, ent._.hypothesis_)

# ...

2025-06-01 20:52:02.612 | WARNING  | edsnlp.pipes.qualifiers.family.family:__init__:171 - You have requested that the pipeline use annotations provided by the `section` pipeline, but it was not set. Skipping that step.
Pipeline components: ['normalizer', 'sentencizer', 'terminology', 'negation', 'family', 'hypothesis']
Available spans: KeysView({'pollutions': []})
Traceback (most recent call last):
  File "test.py", line 33, in <module>
    for ent in doc.spans["terminology"]:
               ~~~~~~~~~^^^^^^^^^^^^^^^
  File "/usr/lib/python3.11/collections/__init__.py", line 1124, in __getitem__
    raise KeyError(key)
KeyError: 'terminology'

My Environment

Debian GNU/Linux 12 (bookworm)
Python: 3.11.2
edsnlp: 0.17.1
spacy: 3.8.6

[aricohen93]

hi @fliot !

You will find your entities in doc.ents.

If you want to add these entities to the spans, you should set a span_setter when defining the pipe (https://aphp.github.io/edsnlp/latest/pipes/core/matcher/)

Example

span_setter=["ents", "ckd"] will add the matches to both doc.ents and doc.spans["ckd"]. It is equivalent to {"ents": True, "ckd": True}.

https://aphp.github.io/edsnlp/latest/pipes/ner/#edsnlp.pipes.base.SpanSetterArg

[fliot]

Hi,
Indeed, I didn't catch the span_setter configuration, great thanks.
Sorry to go further on this basic example, but negation and family pipe seems not applied either.

import edsnlp
import edsnlp.pipes as eds
import pprint

nlp = edsnlp.blank("eds")

nlp.add_pipe(eds.normalizer())
nlp.add_pipe("sentencizer")

terms = {
    "diabete": ["diabète", "diabete", "diabetes"],
    "hta": ["hypertension artérielle", "HTA"]
}

nlp.add_pipe(
    eds.matcher(
        terms=terms,
        attr="NORM",
        term_matcher="exact",
        span_setter="terminology"
    ),
    name="terminology"
)

nlp.add_pipe(eds.negation())
nlp.add_pipe(eds.family())
nlp.add_pipe(eds.hypothesis())

print("Pipeline components:", nlp.pipe_names)

doc = nlp("Pas de diabète. Le père a une hypertension artérielle.")

print("Available spans:", doc.spans.keys())

for ent in doc.spans["terminology"]:
    pprint.pprint([ent.text, ent.label_, ent._.negation_, ent._.family_, ent._.hypothesis_])

output is

2025-06-07 09:08:47.480 | WARNING  | edsnlp.pipes.qualifiers.family.family:__init__:171 - You have requested that the pipeline use annotations provided by the `section` pipeline, but it was not set. Skipping that step.

Pipeline components: ['normalizer', 'sentencizer', 'terminology', 'negation', 'family', 'hypothesis']
Available spans: KeysView({'pollutions': [], 'terminology': [diabète, hypertension artérielle]})
['diabète', 'diabete', None, None, None]
['hypertension artérielle', 'hta', None, None, None]

Where I did expect, negation and family having been found...
Thanks in advance,
Best regards.