Automatic Interpretation Of Loosely Encoded Knowledge