De Informatiesporenmethode (The Information Retrieval Method)

摘要

If one replaces the demand always to find the correct answer by the demand always to find the best possible answer, it is no longer absolutely necessary to introduce into the information retrieval machine comprehension of the language. One may then confine oneself to tracing agreements in meaning between documents without having to know the meaning oneself. The agreement in meaning may be calculated by means of the collection of overlapping trigrams which may be deduced for each text. The direct resemblance between two documents may be expressed in terms of the relative number of common trigrams. But an indirect resemblance may also be defined which may be obtained from the direct resemblance which each of the two documents to be compared has with each of the other documents in the collection. The final resemblance in meaning is a weighted combination of the direct and indirect resemblance. The study aims at exploring the possibility of enabling the user to put his question in natural language and at developing a system which will retrieve the best possible answer from a data bank to a specific question.