Pattern matching for case analysis: a computational definition of closeness

Proposes a conceptually and technically neat method to identify known semantic patterns close to a novel pattern. This occurs in the context of a system to acquire knowledge incrementally from systematically processed expository technical text. This semi-automatic system requires the user to respond to specific multiple-choice questions about the current sentence. The questions are prepared from linguistic elements previously encountered in the text similar to elements in the new sentence. We present a metric to characterize the similarity between semantic case patterns. The computation is based on syntactic indicators of semantic relations and is defined in terms of symbolic pattern matching