Deep Web Entity Identification Method with Unique Constraint

摘要

In pratice, some attributes meet a unique constraint: each entity has a unique value for the attribute. A deep wed entity identification method was presented to solve problems of data error coorection, uniqueness constraiont enforcement, and local data fusion in deep wed data integration. The method transformed the entity identification phrase to a k-partite graph clustering problem, considering both similarity and association of attribute values. Moreover, it performed global record linkage and data fusion simultaneously and could identify incorrect values and differentiate them from correct ones at the beginning. Experimental results demonstrate the high precision and scalability of our method.