Capturing the deep meaning of texts through deduction and inference

One of the main problems in the computer analysis of natural language is understanding sentences beyond a surface level, i.e., making inferences about likely circumstances and drawing plausible conclusions. At the first level, a natural-language-understanding system can answer simple and trivial questions; in order to extend the domain of possible questions that it can answer, the system must make presuppositions and recognize implications that depend on certain events (also called actions). The IBM Rome Scientific Center has developed a prototype system that is able to make inferences about what might be true. This system has been integrated with a text-understanding system (System N), also developed at Rome.