An Exploratory Study of Address Geocoding Techniques with a Focus on Solutioning for the Indian Address Problem
Accurate geocoding of natural language ad- dresses is essential to any industry dealing with logistics services, like the postal service, e-commerce, etc., and there are ongoing research efforts in address geocoding by on-demand service providers. However, currently, existing methods assume the existence of a standard addressing scheme which can be deconstructed and identified.To build automated, scalable and self-learning systems that are capable of analyzing information con- tained in natural language addresses the systems need, along with user input, verification, and updating of learning models. This does not work well in places which do not have a standard addressing scheme or where addresses can not be determined with the help of these addressing schemes due to the massive increase in popu- lation density and urbanization of countries. A number of techniques have been proposed by various authors to provide a solution for address geocoding of natural language addresses that have varying accuracy. There have also been proposed solutions as dynamic models that grow to accommodate changing landscapes.We also discuss relevant issues with these systems such as data collection, evaluation metrics and bench-marking to find accurate location. After going through these proposed solutions and current implementations, we conclude by studying the applicability of these solutions for Indian ad- dresses, their limitations and provide possible directions for future work.