Towards Finding Relational Redescriptions

摘要: This paper introduces relational redescription mining, that is, the task of finding two structurally different patterns describe nearly same set object tuples in a dataset. By extending mining beyond propositional and real-valued attributes, it provides powerful tool to match descriptions concept. As first step towards solving this general task, we introduce an efficient algorithm mines one description given binary A graph is built from frequent path connecting example pairs. Experiments domain explaining kinship terms show approach can produce complex explanations by experts, while being much faster than direct query approach.