We have proposed a recommender system UseLibrary, helping users of library to get proper journal articles, books and other library resources. Profile generation & maintenance and profile exploitation are two main tasks to develop the system. Currently in this system, dataset consists of 790 books and journal articles in PDF and TXT format. System learns the profiles of users and then provides recommendations. Information filtering is required for providing recommendations. Similarity between interest of user and available resources can be computed by different approaches. In this paper, we have compared two approaches, cosine similarity measure and ontology based measure. It is found that use of ontology based measure outperforms cosine similarity measure.