Skip to main content

Table 4 Results on the Quaero Training for PER and NER

From: SIFR annotator: ontology-based semantic annotation of French biomedical text and clinical notes

  Plain Entity Recognition [PER] Normalized Entity Recognition [NER]
P R F1 P R F1 P R F1 P R F1
  EMEA EMEA adapted EMEA EMEA adapted
BSL 64.0 51.7 57.2 63.1 59.3 61.2 49.8 30.9 37.8 48.6 35.1 40.8
DAA 58.3 51.6 54.8 57.5 59.3 58.4 45.0 30.7 36.2 44.0 34.8 38.8
DBP 70.8 56.2 62.6 69.2 64.0 66.7 54.21 31.0 39.4 54.1 35.36 42.8
Avg. 58.7 47.3 51.1 Not Available 33.3 46.0 34.7 Not Available
Med. 73.1 55.9 61.3 19.1 56.5 25.2
  MEDLINE MEDLINE adapted MEDLINE MEDLINE adapted
BSL 57.5 49.0 52.9 55.2 55.8 55.5 44.0 30.5 36.0 43.8 35.5 39.2
DAA 67.9 49.0 56.9 62.2 55.8 60.2 52.9 30.5 38.7 52.7 35.5 42.4
DBP 64.7 54.0 58.9 62.0 61.1 61.5 49.5 30.4 37.6 49.25 35.4 41.2
Avg. 53.3 39.6 44.0 Not Available 32.1 46.1 34.0 Not Available
Med. 64.9 40.0 48.7 29.5 59.0 22.8
  1. Evaluation on both the EMEA and MEDLINE sub corpora for the original Quaero corpus and our adapted Quaero corpus. For the original corpora, we report on the average and median results of the systems participating in CLEF eHealth 2015 Task 1. Values in bold correspond to the best results in each category
\