As mobile devices such as smart phones and tablet PCs are getting popularity, mobile applications have attracted a lot of attentions. Mobile applications have advantages that they can provide information at any time and any place using mobile internet network. However mobile applications suffer from slower response time than desktop applications because of the limited resources of the mobile devices and relatively slow speed of the wireless network. In this paper, we propose a prefetching scheme to improve the response time of the mobile applications. The prefetching is a technique that transfers the next results in advance by predicting the next queries. The prefetching is widely used in web environments. In this paper, we propose to apply the prefetching techniques in a specific domain of the mobile applications; the mobile dictionary application. As prediction methods, we analyze and compare the precedence mining technique and the collaborative filtering and suggest the hybrid method of these two techniques. In addition. we propose a 2 level prefetching scheme in order to minimize network costs while improving the response time. The 2-level prefetching scheme uses a client-side cache and a proxy server-side cache to supplement the loss which are resulted from the client-side cache misses. Through experiments with the syntactic data and the real world data, we show that the proposed techniques can improve the response time of the mobile dictionary application. Although the proposed techniques are suggested in the dictionary domain, our techniques can be applied to other mobile applications with the adequate prediction methods.