Finding relevant and related publications is an important task of researchers’ activities. At Mendeley, we have tens of millions of research articles that we try to recommend to millions of researchers, requiring a large scale solution to this problem. Spark’s implementations of recommender systems have recently attracted much attention. In this presentation, we demonstrate how Spark can be used to generate scientific article recommendations for researchers. We share Mendeley’s experiences of moving from other machine learning libraries to Spark, the challenges that we faced and the solutions that we put in place.
Maya Hristakeva is a Senior Data Scientist at Mendeley working on the next generation of solutions to help researchers connect to their research and to collaborators. Her research interests are in the areas of scalable machine learning, recommender systems and optimization algorithms. Prior to Mendeley, Maya worked as a Researcher at Silicon Valley based startups, Cognitive Match, focused on using machine learning techniques for behavioural targeting and recommendations, and at C8 Medisensors, where she developed algorithms for continuous non-invasive measurement of blood sugar in humans.