LITL at CLEF eHealth2017: automatic classication of death reports

Lydia-Mai Ho-Dac (1) , Cécile Fabre (1) , Anouk Birski (2) , Imane Boudraa (2) , Aline Bourriot (2) , Manon Cassier (2) , Léa Delvenne (2) , Charline Garcia-Gonzalez (2) , Eun-Bee Kang (2) , Elisa Piccinini (2) , Camille Rohrbacher (2) , Aure Séguier (2)

1 CLLE-ERSS - Cognition, Langues, Langage, Ergonomie
2 UT2J - Université Toulouse - Jean Jaurès

This paper describes the participation of a group of students supervised by two teachers to the CLEF eHealth 2017 campaign, task 1. The task involves the classication of death certicates in French and more precisely the labelling of each cause of death with the relevant ICD10 code. The system that performs the automatic coding is based on an information retrieval method using the Solr interface. Two runs were submitted according to whether the system distinguishes cases of multiple causes or not. The best performance was obtained with the system which distinguishes multiple causes, with a precision of 0.61 and a recall of 0.55.

Mots clés: Text classication; biomedical texts; code assignment; cause of death extraction; information retrieval

Domaines: Linguistique

Lydia-Mai Ho-Dac, Cécile Fabre, Anouk Birski, Imane Boudraa, Aline Bourriot, et al.. LITL at CLEF eHealth2017: automatic classication of death reports. CLEF eHealth 2017, Sep 2017, Dublin, Ireland. ⟨hal-01702705⟩