The GEONTO project focuses on interoperability of diverse data related to geographic information. We particularly intend to facilitate applications dealing with such a diversity, from the integration of database schemas to the intuitive query of textual documents, being either technical or common documents. The project originates in the idea that each document has been developed according to a given point of view, which can be seen through the used vocabulary, and which can be more formally reflected in an ontology. Ontologies play a key role for integrating multiple and heterogeneous sources of information. They are formal references containing concepts commonly agreed in a given community. They may be used to describe and consistently integrate data. Because they are a set of reference concepts, they are an efficient gateway for the members of a given community when they attempt to access data annotated by the means of the ontology. Because they are formally expressed, they may be used for automating data integration. The first part of the project consists in building several geographic ontologies reflecting several points of view. In order to do that, several approaches relying on natural language processing techniques will be used. The first one will use lexico-syntactic techniques in order to analyse the textual specifications of several geographic databases produced by IGN-France: new ontologies will be built and existing one will be enriched. The second approach will analyse existing geographical texts of the public library "la médiathèque de Pau" with another technique looking for geographical names and typical words surrounding them. The second part of the project will study the alignment of ontologies built in the previous part. During the last years, many researches have been done for ontology alignment. The idea will be to use and extend them for our case. Several techniques will be used, from the comparison of words describing the concepts to instance matching. An important particularity of the work is to study how to align very heterogeneous ontologies in terms of quality, content or structure. Another important point is to go beyond the ontology alignment: we will develop a methodology to globally compare two ontologies and decide how much they reflect similar points of views, and thus how much they could be merged. The third and last part of the project will take advantage of the results of the first parts for some given applications. The first application is a portal for accessing to numerous textual geographic documents. Built and aligned ontologies will be used to better annotate, query and present the documents. The second application is the matching of diverse geographic database schemas. This task is very important for data users who intend to mix sources of knowledge and for data producers who need to maintain different databases. The difficulty of this task relies on the large heterogeneity of schemas. All the works of this project will be developed and tested on actual data. Real data in the geographic field, more particularly in the topographic domain, will be provided by the French national mapping agency and a public library in the town of Pau for experiments. In France, there are numerous and huge databases in this domain (about one hundred of giga-bytes).