Commonsense Ontologies and the Use of Words in Natural Language

Since the appearance of the ‘Semantic Web’ and the development of ‘RDF’ and ’OWL’, ontologies gained new importance in computer science. Ontological structures can be used to make knowledge available to artificial intelligent systems. But such systems need commonsense knowledge to simulate human reasoning beyond the boundaries given by specific domains. For this purpose commonsense ontologies are employed. However, existing commonsense ontologies, e.g. Cyc [Cy10], were constructed over a lengthy period of time. An interesting proposal to reach this in shorter time with less effort is to expose the structure of commonsense knowledge by analyzing the use of words in natural language. Based on this, a method to automatically gain commonsense ontologies with less effort will be presented. The main point of this method is the automatization. For this purpose data mining techniques are applied and an algorithm to generate the resulting ontology out of the gained data is introduced. 1 Exposing the structure of commonsense knowledge To avert the immense effort 1 for the construction of commonsense ontologies, there is need for new methods to gain these ontological structures. Saba [Sa07] proposes to expose – and not newly develop – the structure of commonsense knowledge by analyzing the use of words in natural language. According to this, natural language should guide the process of gaining a commonsense ontology. 2 Four steps toward an ontological structure Asking a child, if it makes sense to say ‘a dog barks’ or ‘a book barks’, one gets an explicit positive or negative answer. On this simple base – the view of a child – the analysis of words is executed. 1 Knowledge was fed over 20 years into the knowledge base of Cyc [Cy10] until the system was able to learn by itself [Wi05].