Elementization of Thai postal addresses: A hybrid approach

Postal addresses are common data among various databases. However, address structures may be defined differently, especially in Thailand, where national data standards have not been established yet. Many information systems allow users to enter addresses in free-form text; all elements of each address are stored in a single field. Comparing free text addresses in many algorithms such as de-duplication and house-holding is difficult. Hence, to enhance data sharing and integration among organizations, elementization of postal addresses in a standard format is essential. This study develops an algorithm for automatically elementizing Thai postal addresses by using a rule-based approach and the Hidden Markov Model. We evaluate our system on a real-life dataset and yield an accuracy of 97%.