Recommendation systems for research

CINF 63

Marc F. Krellenstein, m.krellenstein@elsevier.com, Elsevier, 30 Corporate Drive, Burlington, MA 01803
As the relevant research literature expands at a rate beyond the ability of any one person to process, technology to recommend work related to your own interests is a valuable tool to help users discover possibly useful information. Collaborative recommendation systems leverage what similar researchers have viewed or done and can provide excellent and reliable suggestions, though there are some obstacles to their use due to privacy considerations and the need for usage history. Content-driven recommendations, perhaps augmented with chemical structure or other similarity measures, do not leverage other (human) judgments but also don't share privacy and usage history limitations, relying only on the existence of available research content and appropriate statistical and natural language technology. These technologies have a good track record of success for similarity searching and are being enhanced to produce ever-better suggestions for related work of interest.