Collection search engine especially for theses and dissertations in Universitas Indonesia's library Online Public Access Catalog (OPAC) website was considered to be less effective and time-consuming as it rarely brings up a match result. This study aimed to develop a categorization system to support theses and dissertations search engine. To build the mentioned system, this study processed text derived from the abstract of each theses and dissertation using Self-organizing Map, one of the applications of clustering. The abstracts of each thesis and dissertation of Universitas Indonesia in 2005-2015 were processed and this study found 139 categories that can represent theses and dissertations of Universitas Indonesia. MySQL was employed to build the proposed categorization system in the database, while PHP programming was used to create a categorization program which will be integrated into the online catalog. By using this categorization system, data collection became more centralized and structured. Data storage is also becoming more efficient and effective so that information is provided more quickly, easily, and accurately.