Towards a Serendipitous Research Paper Recommender System Using Bisociative Information Networks (BisoNets)

In recent times the rate at which information is being processed and shared through the internet has tremendously increased. Internet users are in need of systems and tools that will help them manage this information overload. Search engines and recommendation systems have been recently adopted to help solve this problem. The aim of this research is to model a spontaneous research paper recommender system that recommends serendipitous research papers from two large normally mismatched information spaces or domains using BisoNets. Set and graph theory methods were employed to model the problem, whereas text mining methodologies were used to develop nodes and links of the BisoNets. Nodes were constructed from keywords, while links between nodes were established through weighting that was determined from the co-occurrence of corresponding keywords in the same title and domain. Preliminary results from the word clouds indicates that there is no obvious relationship between the two domains. The strongest links in the established information networks can be exploited to display associations that can be discovered between the two matrices. Research paper recommender systems exploit these latent relationships to recommend serendipitous articles when Bisociative Knowledge Discovery techniques and methodologies are utilized appropriately.