Text categorization or text classification (TC) has recently received increased research attention from information retrieval and machine learning communities, this focus is driven mostly by the ever growing demand for effective and efficient content-based, document management. In the context of digital library or Web portal application, the problem of text categorization is normally that of classification scheme with a topic hierarchy containing all the pre-defined categories. This paper describes our approach to building the hierarchical text classifier for the experimental CINDI Digital Library . The classification system constructed features a top-to-down, coarse-to-fine categorization procedure. We evaluate our system's performance by experiment on a self-generated corpus of the Computer Science papers archived in ACM DL.