Noisy name datasets

These test datasets focuses on string data, which represent persons’ names. These datasets were generated using NSDGen (Noisy String Data Generator) tool. The total amount of elements obtained in each dataset is calculated as the product of k value and the number of exact duplicates. As noise is introduced, duplicates are no longer exact and that amount is referred from now on as the number of observations per group. We refer to noise when common typos as insertions, deletions, substitutions and/or transpositions of characters in strings are introduced. To introduce such typos in strings, we consider the graph of distances among keys in QWERTY keyboards. Them has been used to evaluate clustering algorithms based on partitions in the ambiguous name problem, record linkage and authority control files, as testing datasets.




CC BY 4.0