From the Use of Words in Natural Language to an Ontological Commonsense Structure

To simulate human reasoning, artificial intelligent systems need background knowledge. This knowledge can be represented by an ontology. For specific domains domain ontologies [1] are used. But human inferences beyond the boundaries given by specific domains can only be imitated by means of commonsense knowledge. However, the generation of commonsense ontologies is a very difficult challenge [1]. Existing commonsense ontologies (e.g. Cyc [2]) were constructed over a lengthy period of time. Hence there is need for other methods for gaining commonsense ontologies in shorter time with less effort. A very interesting proposal is to expose the structure of commonsense knowledge (instead of constructing it with immense effort) by analyzing the use of words in natural language [4].