With the information available through World-Wide-Web becoming overwhelming, e fficient information gathering (IG) tools are necessary. Since the network resou rces are expensive, so IG is a resource-bounded task. The main purpose of this paper is to find an efficient gathering method for specific topic. This paper pr esents methods for predicting page's content without downloading it, designs dif ferent controlling strategies, and defines several kinds of page downloading pri ority measures. An IG system, TH-Gatherer, was built to test the methods, and d ifferent experiments were carried out. Through experiments, it was found that th e content of candidate pages can be predicted approximately without downloading. When the priority based gathering strategy and hybrid measure are used, the gat hering efficiency is four times of that of BFS strategy which is used by many cu rrent IG tools (including crawlers and off-line browsing tools). The method pre sented in this paper is suitable for resource-bounded, specific topic informati on gathering.