Building Ontologies for analyzing data expressed in natural language
Abstract
This chapter deals with the issue of translating natural language representations so that their semantic content can be used by a software system. There exist indeed major differences between the informal representations formulated by geologists in natural language and the fully formalized ontology based representations that can be processed by automatic systems. Two methods for building ontologies from and for textual material exploitation are presented. The goal is the construction of data models or concept maps rather than fully formalized ontologies expressed in some ontology language. The first method is description-oriented. It intends to provide a methodology for constructing domain data models from short textual descriptions. It is focused on the multidisciplinary subject of rock description. The second method is document-oriented. It aims at allowing the construction of semi-formal representations such as hierarchies or concept and properties starting from long textual documents. Approaches which try to exploit grammatically structured sets of words possibly nested within full sentences are presented. This second method is instrumented through a collaborative ontology editor which integrates two tools for automatic analysis of natural language.