Automatic extraction of spatial relations

Abstract

We present a method to extract semantic relations from text corpora in an unsupervised way and use the output as preprocessed material for the building of a virtual environment. The experiments have been carried out in the context of the Flemish OntoBasis project. We have used two small corpora. The first one has been gathered from the Internet and is describing the megalithic ruin of Stonehenge. The other one, considered as a test case, is composed in order to describe the appearance and content of a shop. Using a shallow parser, we select functional relations, such as the syntactic structure subject-verb-object, from which we extract spatial information, in an unsupervised way on the Stonehenge corpus, and using domain knowledge for the shop corpus. We then compare and discuss the results, considering the differences between both corpora and to what extent the size of the corpus and the presence of domain knowledge influence the results. The adequacy of the relations extracted has been evaluated manually