Adaptive name matching in information integration

Privacy-Preserving Record Linkage

“…When fields are numeric then perhaps absolute or euclidian difference is appropriate. When fields are strings such as names and addresses, then string edit distances [21,2] are useful. Such distance measures may be thresholded, i.e., reduced to binary match variables where the flag is "on" whenever the distance falls below some cutoff.…”

“…First some random set of strings is agreed upon by the two parties. Then each party computes the edit distance [2] of his records to each random string. With this in hand, the records may be described by a vector of real numbers in which each component is a distance to a random string.…”

“…The record linkage goal is to determine whether a record from one file corresponds to a record of a second file, in the sense that the records describe the same individual. Winkler and others describe application areas, computational techniques and statistical underpinnings in detail in [19,2,38,39]. The typical purposes of record linkage are:…”

From Web Data to Entities and Back

“…At first sight, the entity search task has a strong similarity with entity linkage techniques [14], also known as data matching [4,9], deduplication [27], resolution [3], merge-purge [13], entity identification [21], and reference reconciliation [10]. Entity Linkage is the process that decides whether two descriptions refer to the same real world entity (see [12] for an overview).…”

Valid Statistical Inference on Automatically Matched Files

“…The record linkage goal is to determine whether a record from one file corresponds to a record of a second file, in the sense that the two records describe the same individual. Winkler and others describe application areas, computational techniques and statistical underpinnings in detail in [1,5,9,10]. The typical purposes of record linkage are:…”