S. Christophe

Institut Geographique National (IGN France) COGIT Laboratory



While Web access to mapping tools become more and more democratic, users face difficulties to conceive personalized and effective legends adapted to their taste and needs. Firstly, they seldom have technical and theoretical skills to design efficient maps. Moreover, they may lack creative know-how to draw innovative maps. Lastly, the creation of a legend via a standard graphic interface may require time. Thus it is relevant to assist users in creating innovating and customised legends. This requires methods to facilitate the creation process and to ensure the semiological quality of map.

Our paper deals with the design and implementation of such methods. Our objective is to provide users with an interface to create innovating legends that both fit their taste and needs and complies with cartographic rules.

A first important choice has been to build a dialogue interface and not a classical query or browsing interface. Dialogue techniques are actually required to represent the user needs and to find a compromise solution between those needs and what is cartographically correct.


An important result we rely on is the work of [Hubert]. He proposed a mechanism of man-machine dialogue based on map samples in order to converge towards a satisfying parameterisation in generalisation algorithm. We re-use his proposal: as users could face problems with mapping semantics, the dialog must be based on analogies, i.e. on map samples.

Our challenge is to define communication acts and associated methods to manipulate samples to select them, to build new ones. The first communication act of the system is to build and to propose to the user a selection of map samples. Then the user should express statement like I dont like the colour or I like this sample. The system iterates a proposal which is more and more relevant and more or less varied depending of the dialogue stage.

We propose a model of legend made of themes, objects, relations and colours, which are halfway goals to be reached during the dialogue. In that way we propose four associated strategies: a colour palette to define, theme by theme to process, significant objects to choose, a map sample to refine. Whatever order the user chooses to treat of this strategies, the ultimate goal is reached when the intermediates are too. Lastly we consider that the legend is satisfying when our model of legend is entirely completed.


Hubert F., Ruas A., 2003, A method based on samples to capture user needs for generalisation, 5th Workshop on Progress in Automated Map Generalization, Paris, France, 28-30 April 2003.