Towards Cross Language Morphologic Negation Identification in Electronic Health Records

The current paper presents an approach for analyzing the Electronic Health Records (EHRs) with the goal of automatically identifying morphologic negation such that swapping the truth values of concepts introduced by negation does not interfere with understanding the medical discourse. To identify morphologic negation we propose the RoPreNex strategy that represents the adaptation of our PreNex approach to the Romanian language [1]. We evaluate our proposed solution on the MTsamples [2] dataset. The results we obtained are promising and ensure a reliable negation identification approach in medical documents. We report precision of 92.62 % and recall of 93.60 % in case of the morphologic negation identification for the source language and an overall performance in the morphologic negation identification of 77.78 % precision and 80.77 % recall in case of the target language.