Information Extraction Techniques for Postal Address Standardization

The unique frames of reference of humans result in various definitions of the same details. They develop addresses of same places in different ways, which might result in inconsistent format of addresses ultimately leading to misapprehensions. A major motivation for standardization of the addresses is cleansing of addresses in data warehouses. Since almost every organization deals with a variety of addresses of its customers and employees therefore, a consistent format of addresses can ensure better knowledge of the organization about its customers. This paper presents various information extraction techniques which can also be used in address standardization. It focuses on a statistical model, hidden Markov model (HMM), and two rule-based methods, RAPIER and GRID that extract information from free text. The paper also discusses some personal experience for address standardization