Collaborative ranking and collaborative clustering

Ranking and clustering are two important problems in machine learning and have wide applications in Natural Language Processing (NLP). A ranking problem is typically formulated as ranking a collection of candidate "objects" with respect to a "query" while a clustering problem is formulated as organizing a set of instances into groups such that members in each group share some similarity while members across groups are dissimilar. In this thesis, we introduce collaborative schemes into ranking and clustering problems, and name them as "collaborative ranking" and "collaborative clustering" respectively. Contrast to the tradition non-collaborative schemes, collaborative ranking leverages strengths from multiple query collaborators and ranker collaborators while collaborative clustering leverages strengths from multiple instance collaborators and clusterer collaborators. We select several typical NLP problems as our case studies including entity linking, document clustering and name entity clustering.