Today's natural language technology is inadequate for applications involving untrained public users. For example, no existing system is powerful enough to sell travel services without human assistance. Involved is a design for a general deductive system and underlying representation language integrated with the syntactic processing methods developed earlier. The deductive system will monitor the dialogue between the machine and the user, deciding the meaning of the user's statements, why they were said and how to respond. The representation language expresses both the meaning of the user's statements and the rules for reasoning about them. The innovation lies in using rule-based deduction in the context of a representation language epistemologically adequate for Natural Language applications. It is expected that the research will remove the remaining "technically risky" roadblocks to a natural language system robust enough to be used by the public without prior training.
