Finding Bieber: On removing duplicates from a set of documents

Convert to tfidf space and search for the documents with the highest correlations (using, i.e., a clustering algorithm).