A geocoding algorithm for natural language address with neighborhood properties

Geocoding is widely used in GIS and ITS systems, but the most current geocoding systems have a shortcoming i.e. they cannot deal with some natural language addresses which commonly exist in our oral communication and some other address resources. Thus, geocoding with natural language address becomes a hot spot in GIS geocoding research. This paper, first, put forward a method of determining the neighborhood range of reference entity in POI Address, based on our former research results about modeling neighborhood ranges of different kind of Geographic entity, such as the single and multiple line and face shape entity. Second, based on the model above, an algorithm is put forward to geocoding the natural language address with neighborhood properties. Finally, a prototype system is developed to apply this algorithm. It is concluded that the algorithm has the practical value, because it faithfully fits the commonsense of human beings about the neighborhood concept in natural language address. Furthermore, it also can be applied easily by virtue of its simple principles.