Seminář: Using weighted centroids of word embeddings and top-k documents re-ranking for paper/project-reviewer recommendation

In this talk we present different approaches to content-based recommender systems that are based on centroids of word embeddings (word2vec in particular). This work is motivated by the paper/project-reviewer assignment problem and it will be demonstrated on the medical domain. We will demonstrate a system based on a known approach with TF-IDF centroid weighting + word-mover’s distance re-ranking and also propose a novel method that uses weighting by closeness centrality in the word2vec word similarity graph.