Integrating Graph-Based Document Recommendation in Digital Libraries of Theses Collection

Abstract

Manuscripts in university local collection such as theses and dissertations are very useful resources for researches and literature reviews. The implementation of full-text search and content recommendation within these collections is limited due to local resources. This work aims to enhance document searching capability of local manuscript collections by implementing document index from abstract and developing graph-based document similarity algorithm for related reading recommendation. The evaluation shows that graph-based ranking performed acceptable precision rate at first 2 recommendation items. To enhance the performance of content-based recommendation system, graph-based filtering can be used to reduce search space for traditional document similarity algorithm. We also developed the prototype of graph-based filtering conjunction with document similarity algorithm in our local university theses and dissertation collection.