Approximate String Matching for Multiple-Attribute, Large-Scale Customer Address Databases

摘要

The default pattern matching capabilities in today's RDBMS are generally unable to cope with errors and variations that may exist in stored textual information. In this paper, we present SKIPPER, a simple search methodology that allows approximate string matching on multiple-attribute, large-scale customer address information for the Credit Collection industry. The proposed solution relies on the edit distance error model and the q-gram string filtering technique. We present an algorithm that integrates the methodology with existing RDBMS through SQL-based stored procedures.