FSS-WASTL: Interactive Knowledge Acquisition for a Semantic Lexicon

The following paper describes FSS-WASTL, a system for the acquisition of semantic knowledge within XTRA — a natural language access system to expert systems. Starting from user-supplied example utterances, hypotheses for lexical entries are induced, which then are refined top down through interactive classification. The refinement process of FSS-WASTL is a variant of the well-known KLONE classification procedure and is an extension of T. Finin's interactive classifier. The integration of XTRA's natural language generator, which is a highlight of FSS-WASTL, allows for the paraphrasing of the user's input for clarification purposes. FSS-WASTL is compared with other research systems. It is argued that this approach can be transfered to applications beyond the natural language domain.