The work in this repository had been presented at the SWAT4HCLS conference 2023, Basel, Switzerlands on February 15, 2023.
The presentation is available here and paper will be available soon at the inproceedings of the conference.
The sample sentence:
sample_sentence_en = '''Lauterbach: Omicron is not suitable as a vaccine substitute Federal Health Minister Karl Lauterbach refers to a study from South Africa on Twitter, according to which an Omicron infection in unvaccinated people hardly protects against a disease with the delta variant .'''
model = SequenceTagger.load(model_path)
sentence = Sentence(sample_sentence_en)
model.predict(sentence)
for entity in sentence.get_spans("ner"):
print(entity)Output:
Span[0:1]: "Lauterbach" → PERSON (0.9999)
Span[2:3]: "Omicron" → CORONAVIRUS (1.0)
Span[8:9]: "vaccine" → PRODUCT (0.9746)
Span[13:15]: "Karl Lauterbach" → PERSON (0.9997)
Span[20:22]: "South Africa" → GPE (0.9675)
Span[23:24]: "Twitter" → ORG (0.9986)
Span[29:31]: "Omicron infection" → DISEASE_OR_SYNDROME (0.9965)
Span[32:34]: "unvaccinated people" → GROUP (0.998)
Span[38:39]: "disease" → DISEASE_OR_SYNDROME (0.851)
Span[41:43]: "delta variant" → CORONAVIRUS (0.9608)
sentence = """ How many people want to use the app?
2463.74, According to the ARD Germany trend from June, 42 percent of those surveyed would use such a warning app on their own smartphone, while 39 percent would not."""Output:
Span[2:3]: "people" → GROUP (1.0)
Span[9:10]: "2463.74" → CARDINAL (0.9356)
Span[14:15]: "ARD" → ORG (0.9996)
Span[15:16]: "Germany" → GPE (0.9995)
Span[18:19]: "June" → DATE (0.9871)
Span[20:22]: "42 percent" → PERCENT (0.9954)
Span[37:39]: "39 percent" → PERCENT (1.0)
The usage is available here
The NER model is here