In this paper, we propose an iterative semi-automatic approach for linking textual arguments of relations to their semantic form using rules. Textual arguments are completely decomposed – every word is considered. They are composed back into semantic form using functions, which bring additional semantic information. The process starts with an initial set of seed rules, which can be obtained automatically. In each iteration, the user constructs new rules using the recommendations, which are calculated based on the frequency statistics of unlinked textual arguments. Our approach was tested on extraction of roles that people have in organizations. The results show that only 31 human crafted rules are needed to link more than 3400 additional arguments. We also show that combining rules have positive effects. The number of linked arguments grows super-linearly with respect to the number of patterns.