Since online system evolution requires efficient service selection to meet with the high dynamics demand of open system, this paper proposes a reputation-based recommender discovery approach. It qualifies trust relationships in different recommendation contexts via a relative factor, divides the Web of trust into personalized trust networks by applying a segment algorithm and finally locates recommenders with high reputation through trust opinion iteration among users. Simulation results show that the suggested approach in this paper helps to reduce the cost in information collection as well as improve the efficiency and precision of service selection results.