Measuring lexical similarity methods for textual mapping in nursing diagnoses in Spanish and SNOMED-CT

One of the current problems in the health domain is the reuse and sharing the clinical information between different professionals, as they are written in natural language using specific terminologies. To overcome this issue it is necessary to use a common terminology, like SNOMED-CT, allowing an information reuse that offers the health professionals the quickest access to quality information. In order to use this terminology all the other terminologies have to be mapped to it. One solution to perform that mapping is using a lexical similarity approach. In this paper we analyze the appropriateness of 15 lexical similarity methods for mapping a set of NANDA-I labels to a set of SMOED-CT descriptions in Spanish. Our aim is to establish how to choose the best algorithm in this domain, from the recall and the precision point of view. After running six different tests, we have established that the three best algorithms where those that maximize the recall, because they always return the best solution.