Uma metodologia para construção de geo-ontologias

Abstract

Existing textual information is knowledge-rich and grows exponentially in time. To be useful to systems performing some type of intelligent processing, this knowledge needs to be automatically handled. Ontologies have emerged as the preferred unifying semantic representation of knowledge in texts and databases. However, the construction of ontologies requires the level-headed use of techniques of Natural Language Processing (NLP). This thesis proposes an information extraction and integration methodology for building geo-ontologies from knowledge extracted from multiple sources. Initially, we collect geographic data from selected information sources which are integrated into a geographic knowledge base. This knowledge is then enriched from texts aided by a geographic information extraction and integration system. The results show that the geo-ontologies generated using the proposed methodology are useful to several applications. The geographic information extraction system reaches results at the same level of state of art systems in the task of recognition of places and its relationships in Portuguese