Adaptive duplicate detection using learnable string similarity measures