Using Spatial Image Schemata in the characterization of spatial relations

Abstract. Spatially describing objects or environments is a natural and everyday task present in the daily lives of individuals. To portray the existing relationships between present elements of the scene described, it is necessary to use terms known as spatial relations. Frequently, such descriptions are performed using Natural Language (NL), both spoken and written. The existence of a constant interaction of human with the environment makes NL rich in terms that characterize space, resulting in a diversity of such words. The variety of terms used as spatial relations makes it difficult to implement spatial localization systems that use NL. Therefore, to understand how spatial relations are used, the present article aimed to detect and categorize such spatial relations. For this, an experiment of a spatial description of environments unknown by users was performed. The volunteers were native speakers of Brazilian Portuguese language and from the spatial descriptions were obtained locative expressions that allowed the definition and categorization of spatial relations using the Spatial Image Schemata. The results obtained demonstrated an attempt to understand the spatial relations used in spatial descriptions. In the future, we aim to define a set of spatial relations representative of the words used in a spatial description task.