High-Value Token-Blocking: Efficient Blocking Method for Record Linkage

Data integration is an important component of Big analytics. One the key challenges in data record linkage, that is, matching records represent same real-world entity. Because computational costs, methods referred to as blocking are employed a part linkage pipeline order reduce number comparisons among records. In past decade, range techniques have been proposed. Real-world applications require approaches can handle heterogeneous sources and do not rely on labelled data. We propose high-value token-blocking (HVTB), simple efficient approach for unsupervised schema-agnostic, based crafted use Term Frequency-Inverse Document Frequency. compare HVTB with multiple over datasets, including novel unstructured dataset composed titles abstracts scientific papers. thoroughly discuss results terms accuracy, resources, different characteristics datasets The simplicity yields fast computations does harm its accuracy when compared existing approaches. It shown be significantly superior other methods, suggesting simpler should considered before resorting more sophisticated methods.