Capturing Information on Behaviour with the RADD-NLI: A Linguistic and Knowledge-Based Approach

In this paper we illustrate how information on behaviour can be captured from natural language with the RADD-NLI. The behavioural information contained in verbal descriptions of processes is transformed into an abstract process model of characteristic processes. This abstract process model is then integrated into the knowledge base of the RADD-NLI. Although taken from a speciic domain this model is abstract enough to form the basis for further applications with synonym relations. Processes coming up in the user dialogue will be matched against the knowledge base. A dialogue control is activated if a mis-match occurs or for the detail discussion of captured processes with the user. Linguistic knowledge, especially antonymous verb pairs, is considered to decide whether a captured process will be integrated into the structural database design and/or the process list of the design.