Correcting and Standardizing Crude Drug Names in Traditional Medicine Formulae by Ensemble of String Matching Techniques

Abstract

Common problems of representing crude drug names in traditional herbal formulae are spelling errors, grammatical variants, synonyms and various formats. In order to make these names more obvious and useful, correcting and standardizing of these names should be applied. In this work, crude drug names in various forms were corrected and standardized by string matching techniques. A set of experiments were done using crude drug names from a database of registered traditional medicines in Thai Food and Drug Administration as the test set. Two well-known algorithms, i.e., similar text and Levenshtein were investigated. However, the results from each algorithm indicated that crude drug names in the test set were moderately matched with those of the standard set. To increase performance of these single algorithms, the ensemble algorithm was proposed. From the results, the ensemble algorithm outperforms single algorithms to match crude drug names, especially crude drug names with the modifier that have no significant meaning.