How to help reusers retrieve components efficiently and conveniently is critical to the success of the component-based software development (CBSD). In the literature, many research efforts have been put into the improvement of component retrieval mechanisms. Although many different retrieval methods have been proposed, nowadays the component description text is still an important information source for retrieval. Unfortunately, the descriptions of components often contain improper or even noisy information which could deteriorate the effectiveness of the retrieval mechanism. To alleviate the problem, in this paper, we propose an approach which can refine the component description by mining user query logs. The key idea of our approach is to extract proper information from the user query logs to amend the description of the component. Besides, logs from similar components are also taken into consideration. To evaluate the effectiveness of our approach, we performed an experimental study on the JDK. The experimental results demonstrated an improvement in terms of retrieval effectiveness.