Extracting non-taxonomic relations is one of the important tasks in the construction of ontology from the text. Most of current methods on identification and extraction of non-taxonomic relations is based on predicate representing relationships between two concepts, namely the relation between subject and object that occurs in a sentence. However, the number of relations that has been identified does not properly represent the domain as the methods only identify a portion of the total relations from domain texts. In this paper, we present a method that increases the number of relations extracted and thus properly represent the domain. In this method, all potential relations are first generated and then less significant ones, based on their frequency, are removed. The method has been tested on a collection of texts that described electronic voting machine and the result is encouraging. 2011 IEEE.