Recommending users to follow based on user taste recommending users to follow based on user taste

Online content curation social networks are an increasingly popular type of social networks where users pin items they find on the Web into categorical boards defined by users (hence the "curation"). In this paper, we study the problem of recommending users to follow on such social networks. Different from existing friendship-oriented or content-oriented user recommendation approaches, we design a recommendation scheme by collaborative filtering based on homophily for users' tastes. We propose to discover homophily for users' tastes from their "repin paths". Specifically we measure the similarity between repin paths using the Levenshtein Distance. The similarity and the number of common users for repin paths are combined to compute the homophily for users' tastes. Experiments on a content curation social network show that our recommendation algorithm of collaborative filtering based on user taste homophily performs better than user popularity based recommendation.