Automatic extraction of semantic relations from french clinical reports: a pattern based approach
Aicha Ghoulam, Fatiha Barigou and Ghalem Belalem

Information extraction is an automatic process that aims to analyse texts written in natural language to extract structured and useful information such as named entities and semantic relations linking these entities. In this paper, we present and evaluate linguistic pattern based approach for the extraction and annotation of semantic relations that exist between two medical entities from French clinical reports. Our proposed approach to extract these relations is based on two steps; in the first step, we extract medical entities from sentences and determine their categories. In the second step, we extract semantic relations between the extracted entities using linguistic patterns relying on local grammar. We evaluated the semantic relation extraction approach using two different methods that are based on the type of recognition used for the medical entities. The first evaluation used a manual recognition of medical entity, the relation extraction yielded 75.65% precision and 46.1% recall. The second evaluation used the automated medical entity recognition the system achieved a precision of 56.25% and a recall of 33.05%.