Quasi comprehension of natural language simulated by means of information traces

Abstract Good information storage and retrieval systems should not only meet the looseness of human memory but also the looseness of natural language. Our information trace method allows for the creation of mechanical systems that combine looseness with indulgence. In accordance with this method, information records consisting of pieces of natural language are attached to each other, leaving fuzzy semantic traces behind in the form of sets of three-letter-strings which have been derived from the original texts. The system calculates an “importance value” for each possible answer to a user query. Queries also may be given in the form of a simple piece of natural language. Answers are communicated to the user, if he so wishes, in descending order of “importance value” as calculated by the system.