A Multilayered Approach To The Handling Of Word Formation

The treatment of word formations has until recently been a neglected topic in natural language AI research. This paper proposes a multilayered approach to word formation which treats derivatives and compounds on several different levels of processing within a natural language dialogue system. Analysis and generation strategies being developed for the dialogue system HAM-ANS are described. Identification of word formations, semantic interpretation, and evaluation in the context of a dialogue are the main • levels of analysis on which the system successively attempts to infer the implicit relations between word formation components. Generation of word formations is viewed as a process comparable to the generation of elliptical utterances.