Scalability of Piecewise Synonym Identification in Integration of SNOMED into the UMLS

Synonym identification during source terminology integration into the Unified Medical Language System (UMLS) is a labor-intensive task needed for every new release of the source. The piecewise synonym (PWS) methodology was previously used for the integration of a small source. The goal of this paper is to determine whether the piecewise synonym methodology with two control parameters scales to a much larger terminology (a subset of SNOMED CT), the control parameters are necessary to make the methodology viable, and the control parameters lead to any loss of matching results. Additional methods for limiting the size of the dictionary used in the PWS generation methodology are used. The authors’ methodology discovered 41% of concepts not found by string matching. The necessity and effectiveness of the control parameters were confirmed. Furthermore, when comparing the results of experiments with and without control parameters, no matches were lost.