Optimizing a Near-duplicate Document Detection System with SIMD Technologies