2019/04/25 by Ivan Lerner, Lerner, Ivan, Nicolás Paris +3
Biochemistry, Genetics and Molecular Biology · Computer Science · #Biomedical Text Mining and Ontologies #Computation and Language (cs.CL) #FOS: Computer and information sciences #Natural Language Processing Techniques #Topic Modeling
paper · pdf · doi:10.48550/arxiv.1904.11473
openalex publication_date 2019/04/25 · openalex created_date 2025/10/10 · openalex updated_date 2026/07/28
We aimed to enhance the performance of a supervised model for clinical\nnamed-entity recognition (NER) using medical terminologies. In order to\nevaluate our system in French, we built a corpus for 5 types of clinical\nentities. We used a terminology-based system as baseline, built upon UMLS and\nSNOMED. Then, we evaluated a biGRU-CRF, and an hybrid system using the\nprediction of the terminology-based system as feature for the biGRU-CRF. In\nEnglish, we evaluated the NER systems on the i2b2-2009 Medication Challenge for\nDrug name recognition, which contained 8,573 entities for 268 documents. In\nFrench, we built APcNER, a corpus of 147 documents annotated for 5 entities\n(drug name, sign or symptom, disease or disorder, diagnostic procedure or lab\ntest and therapeutic procedure). We evaluated each NER systems using exact and\npartial match definition of F-measure for NER. The APcNER contains 4,837\nentities which took 28 hours to annotate, the inter-annotator agreement was\nacceptable for Drug name in exact match (85%) and acceptable for other entity\ntypes in non-exact match (>70%). For drug name recognition on both i2b2-2009\nand APcNER, the biGRU-CRF performed better that the terminology-based system,\nwith an exact-match F-measure of 91.1% versus 73% and 81.9% versus 75%\nrespectively. Moreover, the hybrid system outperformed the biGRU-CRF, with an\nexact-match F-measure of 92.2% versus 91.1% (i2b2-2009) and 88.4% versus 81.9%\n(APcNER). On APcNER corpus, the micro-average F-measure of the hybrid system on\nthe 5 entities was 69.5% in exact match, and 84.1% in non-exact match. APcNER\nis a French corpus for clinical-NER of five type of entities which covers a\nlarge variety of document types. Extending supervised model with terminology\nallowed for an easy performance gain, especially in low regimes of entities,\nand established near state of the art results on the i2b2-2009 corpus.\n