Context: In general, applications that interact with geographic information systems work correctly when users use previously geo-coded points. However, problems occur if any of the selected points are not geo-coded. In the particular case of the calculation of routes in which one of the points is not geo-co­dified, it is not possible to calculate the route since one or both points are not identifiable; Therefore, it is necessary to approximate the points that the user has selected to the nearest geo-located point. On the other hand, the case is often presented in which users wish to interact with addresses in common language, regardless of whether the application internally does so based on geographical coordinate points.Method: One of the possible solutions to the pre­vious problem is to propose the construction of an inverse locator that would be responsible for making the translations of geographical coordinates to natu­ral directions. For the construction of said locator the following steps are proposed: Firstly, the specification of the type of application in which the inverse locator will be implemented; second, the analysis of the ele­ments that must be taken into account for the refine­ment of the architecture; third, the creation of a first version of the architecture; and finally the creation of a second version of the architecture that would co­rrespond to the final version of the architecture.Results: The end result is the creation of an archi­tecture that implements a reverse locator that allows this type of geographic applications to solve the pro­blems raised in the context part of the abstract.Conclusions: As a result of the research, we present the conditions under which the inverse locator cons­truction is valid, and the possible problems that will arise if the architecture is not correctly applied. The paper also presents conclusions related to potential improvements that allow to solve the problem inclu­ding adaptive and intelligent locations.