Adaptive text retrieval system

This paper deals with using of the neural networks in information retrieval. There are studied and described some already implemented information retrieval systems based on neural networks and summarized possible advantages and disadvantages of this access in this paper. On the basis of them is designed the own adaptive system called ADIS, built on layered network and the backpropagation algorithm. System simulates user's imaginations using term association acquired from relevance feedback. Design contains discussion of different variants, options and parameters of the system. The thesis also includes implementation and practical tests on collection containing about thousands documents