Using Named Entities from outside Spacy

[jenojp]

Thank you to Markus Hartenfeller for providing this gist solution to importing named entities from outside spacy.

import spacy
from negspacy.negation import Negex
from negspacy.termsets import termset



input_text = "There is no evidence of pneumonia."

nlp = spacy.blank("en")             # that's sufficient and much more lightweight
nlp.add_pipe('sentencizer')         # add a single pipeline component (sentence boundaries are needed by ConText)

print('pipeline components: ', nlp.pipe_names)


doc = nlp(input_text)               # apply pipeline to text

print("Entities before entity infusion: ", doc.ents)  # --> no entities detected at this point because the pipeline contains no NER component


# define entities based on char offset coordinates as returned by BERN2, as well as the returned entity types
cust_ent = doc.char_span(24,33 ,label="DISEASE") # fits to example "There is no evidence of pneumonia."
doc.set_ents([cust_ent], default="unmodified")

ents = [(e.text, e.start, e.end, e.label_) for e in doc.ents]
print('After adding entities', ents)




# now apply NegSpacy to custom document
default_ts = termset("en_clinical").get_patterns()

negex = Negex(nlp, name='testnegspacy', extension_name="is_negex", ent_types = ["DISEASE"], chunk_prefix= list(), neg_termset=default_ts)  # here we could specify our own ruleset by giving a path to a json fle containing our custom rules, default is under /Users/Markus/anaconda3/envs/nlp/lib/python3.10/site-packages/resources/context_rules.json

negex(doc)        #apply ConText ruleset to our doc, defined by BERN2's entities


ents_2 = [(e.text, e.start, e.end, e.label_, e._.is_negex) for e in doc.ents]  # has to match the extension flag ._.is_x as defines above when we created the negex object
print('After Negex semantic modifier detection:', ents_2)