What's up? research

Abstract

Nowadays, a lot of information is available on the internet, causing users to be overwhelmed by all the provided information, and thus leading to information overload. A problem which arises due to this issue is the difficulty for academic researchers to find conferences on the internet, or through received messages, which interest them enough to publish papers to and/or attend. This leads to a COnference REcommender (CORE), which makes use of conference Call For Papers (CFPs) to recommend conferences to researchers. The approach is to use CFPs as a source of information about conferences, since they include properties such as the series, title, dates and location of the conference to be held. Fetching this information and storing it inside a graph database creates the conferences layer. Users' information is collected from online sources which present their publications, creating the second layer of the graph, the user layer. The two layers are connected using content-based (through conference series and research topics) and collaborative-filtering-based (through user co-authors) links. Both content-based and collaborative ltering-based are implemented in CORE, along with a hybrid approach which combines the two. The best result was obtained when evaluating using the hybrid approach, giving a recall of 0.311, a precision of 0.248 and an f1 of 0.257.